Active learning is a form of learning in which teaching strives to involve students in the learning process more directly than in other methods.
Bonwell (1991) 'states that in active learning, students participate in the process and students participate when they are doing something besides passively listening.' (Weltman, p. 7) Active learning is 'a method of learning in which students are actively or experientially involved in the learning process and where there are different levels of active learning, depending on student involvement. (Bonwell & Eison 1991). In the Association for the Study of Higher Education (ASHE) report the authors discuss a variety of methodologies for promoting 'active learning'. They cite literature that indicates that to learn, students must do more than just listen: They must read, write, discuss, or be engaged in solving problems. It relates to the three learning domains referred to as knowledge, skills and attitudes (KSA), and that this taxonomy of learning behaviours can be thought of as 'the goals of the learning process'.[1] In particular, students must engage in such higher-order thinking tasks as analysis, synthesis, and evaluation.[2]Active learning engages students in two aspects – doing things and thinking about the things they are doing.
- 1Nature of active learning
- 2Science of learning in active learning
- 7References
Nature of active learning[edit]
There are a wide range of alternatives for the term 'active learning' like learning through play, technology-based learning, activity-based learning, group work, project method, etc. the underlying factor behind these are some significant qualities and characteristics of active learning. Active learning is the opposite of passive learning; it is learner-centered, not teacher-centered, and requires more than just listening; the active participation of each and every student is a necessary aspect in active learning. Students must be doing things and simultaneously think about the work done and the purpose behind it so that they can enhance their higher order thinking capabilities. Many research studies[by whom?] have proven that active learning as a strategy has promoted achievement levels and some others[who?] say that content mastery is possible through active learning strategies. However, some students as well as teachers find it difficult to adapt to the new learning technique.[3] Active learning should transform students from passive listeners to active participants and helps students understand the subject through inquiry, gathering and analyzing data to solving higher order cognitive problems. There is intensive use of scientific and quantitative literacy across the curriculum and technology-based learning is also in high demand in concern with active learning.[4] Barnes (1989)[5][6] suggested principles of active learning:
- Purposive: the relevance of the task to the students' concerns.
- Reflective: students' reflection on the meaning of what is learned.
- Negotiated: negotiation of goals and methods of learning between students and teachers.
- Critical: students appreciate different ways and means of learning the content.
- Complex: students compare learning tasks with complexities existing in real life and making reflective analysis.
- Situation-driven: the need of the situation is considered in order to establish learning tasks.
- Engaged: real life tasks are reflected in the activities conducted for learning.
Active learning requires appropriate learning environments through the implementation of correct strategy. Characteristics of learning environment are:[7][8]
- Aligned with constructivist strategies and evolved from traditional philosophies.
- Promoting research based learning through investigation and contains authentic scholarly content.
- Encouraging leadership skills of the students through self-development activities.
- Creating atmosphere suitable for collaborative learning for building knowledgeable learning communities.
- Cultivating a dynamic environment through interdisciplinary learning and generating high-profile activities for a better learning experience.
- Integration of prior with new knowledge to incur a rich structure of knowledge among the students.
- Task-based performance enhancement by giving the students a realistic practical sense of the subject matter learnt in the classroom.
Constructivist framework[edit]
Active learning coordinates with the principles of constructivism which are, cognitive, meta-cognitive, evolving and affective in nature. Studies have shown that immediate results in construction of knowledge is not possible through active learning, the child goes through process of knowledge construction, knowledge recording and knowledge absorption. This process of knowledge construction is dependent on previous knowledge of the learner where the learner is self-aware of the process of cognition and can control and regulate it by themselves.[9] There are several aspects of learning and some of them are:
- Learning through meaningful reception by David Ausubel, he emphasizes the previous knowledge the learner possesses and considers it a key factor in learning.
- Learning through discovery by Jerome Bruner, where students learn through discovery of ideas with the help of situations provided by the teacher.
- Conceptual change: misconceptions takes place as students discover knowledge without any guidance; teachers provide knowledge keeping in mind the common misconceptions about the content and keep an evaluatory check on the knowledge constructed by the students.
- Social Constructivism by Bandura and Vygotsky, collaborative group work within the framework of cognitive strategies like questioning, clarifying, predicting and summarizing.[10]
Science of learning in active learning[edit]
Active learning has been definitively shown to be superior to lectures in promoting both comprehension and memory (Freeman et al., 2014). The reason it is so effective is that it draws on underlying characteristics of how the mind and brain operate during learning. These characteristics have been documented by thousands of empirical studies (e.g., Smith & Kosslyn, 2011) and have been organized into a set of principles. Each of these principles can be drawn on by various active learning exercises. They also offer a framework for designing activities that will promote learning; when used systematically, Stephen Kosslyn (2017) notes these principles enable students to “learn effectively—sometimes without even trying to learn.” [11]
The principles of learning[edit]
One way to organize the empirical literature on learning and memory specifies 16 distinct principles, which fall under two umbrella “maxims.” The first maxim, “Think it Through,” includes principles related to paying close attention and thinking deeply about new information. The second, “Make and Use Associations,” focuses on techniques for organizing, storing, and retrieving information.
The principles can be summarized as follows:
Maxim I: Think it Through[11]
1. Evoking deep processing: extending thinking beyond “face value” of information (Craig et al., 2006; Craik & Lockhart, 1972)
2. Using desirable difficulty: ensuring that the activity is neither too easy nor too hard (Bjork, 1988, 1999; VanLehn et al., 2007)
3. Eliciting the generation effect: requiring recall of relevant information (Butler & Roediger, 2007; Roediger & Karpicke, 2006)
4. Engaging in deliberate practice: promoting practice focused on learning from errors (Brown, Roediger, & McDaniel, 2014; Ericsson, Krampe, & Tesch-Romer, 1993)
5. Using interleaving: intermixing different problem types
6. Inducing dual coding: presenting information both verbally and visually (Kosslyn, 1994; Mayer, 2001; Moreno & Valdez, 2005)
7. Evoking emotion: generating feelings to enhance recall (Erk et al., 2003; Levine & Pizarro, 2004; McGaugh, 2003, 2004)
Maxim II: Make and Use Associations[11]
8. Promoting chunking: collecting information into organized units (Brown, Roediger, & McDaniel, 2014; Mayer & Moreno, 2003)
9. Building on prior associations: connecting new information to previously stored information (Bransford, Brown, & Cocking, 2000; Glenberg & Robertson, 1999; Mayer, 2001)
10. Presenting foundational material first: providing basic information as a structural “spine” onto which new information can be attached (Bransford, Brown, & Cocking, 2000; Wandersee, Mintzes, & Novak, 1994)
11. Exploiting appropriate examples: offering examples of the same idea in multiple contexts (Hakel & Halpern, 2005)
12. Relying on principles, not rote: explicitly characterizing the dimensions, factors or mechanisms that underlie a phenomenon (Kozma & Russell, 1997; Bransford, Brown, & Cocking, 2000)
13. Creating associative chaining: sequencing chunks of information into stories (Bower & Clark, 1969; Graeser, Olde, & Klettke, 2002)
14. Using spaced practice: spreading learning out over time (Brown, Roediger, & McDaniel, 2014; Cepeda et al., 2006, 2008; Cull, 2000)
15. Establishing different contexts: associating material with a variety of settings (Hakel & Halpern, 2005; Van Merrienboer et al., 2006)
16. Avoiding interference: incorporating distinctive retrieval cues to avoid confusion (Adams, 1967; Anderson & Neely, 1996)
Active learning typically draws on combinations of these principles. For example, a well-run debate will draw on virtually all, with the exceptions of dual coding, interleaving, and spaced practice. In contrast, passively listening to a lecture rarely draws on any.
Active learning exercises[edit]
Bonwell and Eison (1991) suggested learners work collaboratively, discuss materials while role-playing, debate, engage in case study, take part in cooperative learning, or produce short written exercises, etc. The argument is 'when should active learning exercises be used during instruction?'. Numerous studies have shown that introducing active learning activities (such as simulations, games, contrasting cases, labs,.) before, rather than after lectures or readings, results in deeper learning, understanding, and transfer.[12][13][14][15][16][17][18][19] The degree of instructor guidance students need while being 'active' may vary according to the task and its place in a teaching unit.In an active learning environment learners are immersed in experiences within which they engage in meaning-making inquiry, action, imagination, invention, interaction, hypothesizing and personal reflection (Cranton 2012).
Examples of 'active learning' activities include
- A class discussion may be held in person or in an online environment. Discussions can be conducted with any class size, although it is typically more effective in smaller group settings. This environment allows for instructor guidance of the learning experience. Discussion requires the learners to think critically on the subject matter and use logic to evaluate their and others' positions. As learners are expected to discuss material constructively and intelligently, a discussion is a good follow-up activity given the unit has been sufficiently covered already.[20] Some of the benefits of using discussion as a method of learning are that it helps students explore a diversity of perspectives, it increases intellectual agility, it shows respect for students’ voices and experiences, it develops habits of collaborative learning, it helps students develop skills of synthesis and integration (Brookfield 2005). In addition, by having the teacher actively engage with the students, it allows for them to come to class better prepared and aware of what is taking place in the classroom.[21]
- A think-pair-share activity is when learners take a minute to ponder the previous lesson, later to discuss it with one or more of their peers, finally to share it with the class as part of a formal discussion. It is during this formal discussion that the instructor should clarify misconceptions. However students need a background in the subject matter to converse in a meaningful way. Therefore, a 'think-pair-share' exercise is useful in situations where learners can identify and relate what they already know to others. So preparation is key. Prepare learners with sound instruction before expecting them to discuss it on their own. If properly implemented, it saves instructor time, keeps students prepared, helps students to get more involved in class discussion and participation and provide cumulative assessment of student progress. The 'think-pair-share' method is useful for teachers to hear from all students even those who are quiet in class. This teaching method functions as a great way for all the students in the class to get involved and learn to work together and feel comfortable sharing ideas. It can also help teachers or instructors to observe students and see if they understand the material being discussed.[22] This is not a good strategy to use in large classes because of time and logistical constraints (Bonwell and Eison, 1991). Think-pair-share is helpful for the instructor as it enables organizing content and tracking students on where they are relative to the topic being discussed in class, saves time so that he/she can move to other topics, helps to make the class more interactive, provides opportunities for students to interact with each other (Radhakrishna, Ewing, and Chikthimmah, 2012).
- A learning cell is an effective way for a pair of students to study and learn together. The learning cell was developed by Marcel Goldschmid of the Swiss Federal Institute of Technology in Lausanne (Goldschmid, 1971). A learning cell is a process of learning where two students alternate asking and answering questions on commonly read materials. To prepare for the assignment, the students read the assignment and write down questions that they have about the reading. At the next class meeting, the teacher randomly puts students in pairs. The process begins by designating one student from each group to begin by asking one of their questions to the other. Once the two students discuss the question, the other student ask a question and they alternate accordingly. During this time, the teacher goes from group to group giving feedback and answering questions. This system is also called a student dyad.
- A short written exercise that is often used is the 'one-minute paper.' This is a good way to review materials and provide feedback. However a 'one-minute paper' does not take one minute and for students to concisely summarize it is suggested[who?] that they have at least 10 minutes to work on this exercise. (See also: Quiz#In education.)
- A collaborative learning group is a successful way to learn different material for different classes. It is where you assign students in groups of 3-6 people and they are given an assignment or task to work on together. This assignment could be either to answer a question to present to the entire class or a project. Make sure that the students in the group choose a leader and a note-taker to keep them on track with the process. This is a good example of active learning because it causes the students to review the work that is being required at an earlier time to participate. (McKinney, Kathleen. (2010). Active Learning. Normal, IL. Center for Teaching, Learning & Technology.) To create participation and draw on the wisdom of all the learners the classroom arrangement needs to be flexible seating to allow for the creation of small groups. (Bens, 2005)
- A student debate is an active way for students to learn because they allow students the chance to take a position and gather information to support their view and explain it to others. These debates not only give the student a chance to participate in a fun activity but it also lets them gain some experience with giving a verbal presentation. (McKinney, Kathleen. (2010). Active Learning. Normal, IL. Center for Teaching, Learning & Technology.)
- A reaction to a video is also an example of active learning because most students love to watch movies. The video helps the student to understand what they are learning at the time in an alternative presentation mode. Make sure that the video relates to the topic that they are studying at the moment. Try to include a few questions before you start the video so they pay more attention and notice where to focus at during the video. After the video is complete divide the students either into groups or pairs so that they may discuss what they learned and write a review or reaction to the movie. (McKinney, Kathleen. (2010). Active Learning. Normal, IL. Center for Teaching, Learning & Technology.)
- A small group discussion is also an example of active learning because it allows students to express themselves in the classroom. It is more likely for students to participate in small group discussions than in a normal classroom lecture because they are in a more comfortable setting amongst their peers, and from a sheer numbers perspective, by dividing the students up more students get opportunities to speak out. There are so many different ways a teacher can implement small group discussion in to the class, such as making a game out of it, a competition, or an assignment. Statistics show that small group discussions is more beneficial to students than large group discussions when it comes to participation, expressing thoughts, understanding issues, applying issues, and overall status of knowledge.[23]
- Just-in-time teaching promotes active learning by using pre-class questions to create common ground among students and teachers before the class period begins. These warmup exercises are generally open ended questions designed to encourage students to prepare for class and to elicit student's thoughts on learning goals.
- A class game is also considered an energetic way to learn because it not only helps the students to review the course material before a big exam but it helps them to enjoy learning about a topic. Different games such as Jeopardy! and crossword puzzles always seem to get the students' minds going. (McKinney, Kathleen. (2010). Active Learning. Normal, IL. Center for Teaching, Learning & Technology.)
- Learning by teaching is also an example of active learning because students actively research a topic and prepare the information so that they can teach it to the class. This helps students learn their own topic even better and sometimes students learn and communicate better with their peers than their teachers.
- Gallery Walk is also an example of active learning where students in groups move around the classroom or workshop actively engaging in discussions and contributing to other groups and finally constructing knowledge on a topic and sharing it.
Use of technology[edit]
To have active learning experience, use of technology tools and multimedia helps enhance the atmosphere of the classroom. Each student actively engages in the learning process. Using movies and games the teacher can make the experience more effective. The theoretical foundation of this learning process are :
- Flow: Flow is a concept to enhance the focus level of the student as each and every individual becomes aware and completely involved in the learning atmosphere. In accordance with one's own capability and potential, through self-awareness, students perform the task at hand. The first methodology to measure flow was Csikszentmihalyi's Experience Sampling (ESM).
- Learning Styles: Acquiring knowledge through one's own technique is called learning style. Learning occurs in accordance with one's own potential as every child is different and has potential in different areas. It caters to all kinds of learners: visual, kinaesthetic, cognitive and affective.[dubious]
- Locus of Control: Ones with high internal locus of control believe that every situation or event is attributable to their resources and behavior. Ones with high external locus of control believe that nothing is under their control.
- Intrinsic Motivation: Intrinsic motivation is a factor that deals with self-perception about the task at hand. Interest, attitude, and results depend on the self-perception of the given activity.[24]
Research evidence[edit]
Numerous studies have shown evidence to support active learning, given adequate prior instruction.
A meta-analysis of 225 studies comparing traditional lecture to active learning in university math, science, and engineering courses found that active learning reduces failure rates from 32% to 21%, and increases student performance on course assessments and concept inventories by 0.47 standard deviations. Because the findings were so robust with regard to study methodology, extent of controls, and subject matter, the National Academy of Science publication suggests that it might be unethical to continue to use traditional lecture approach as a control group in such studies. The largest positive effects were seen in class sizes under 50 students and among students under-represented in STEM fields.[25]
Richard Hake (1998) reviewed data from over 6000 physics students in 62 introductory physics courses and found that students in classes that utilized active learning and interactive engagement techniques improved 25 percent points, achieving an average gain of 48% on a standard test of physics conceptual knowledge, the Force Concept Inventory, compared to a gain of 23% for students in traditional, lecture-based courses.[26]
Similarly, Hoellwarth & Moelter (2011)[27] showed that when instructors switched their physics classes from traditional instruction to active learning, student learning improved 38 percent points, from around 12% to over 50%, as measured by the Force Concept Inventory, which has become the standard measure of student learning in physics courses.
In 'Does Active Learning Work? A Review of the Research', Prince (2004) found that 'there is broad but uneven support for the core elements of active, collaborative,cooperative and problem-based learning' in engineering education.[28]
Michael (2006),[29] in reviewing the applicability of active learning to physiology education, found a 'growing body of research within specific scientific teaching communities that supports and validates the new approaches to teaching that have been adopted.'
In a 2012 report titled 'Engage to Excel',[30] the United States President's Council of Advisors on Science and Technology (PCAST) described how improved teaching methods, including engaging students in active learning, will increase student retention and improve performance in STEM courses. One study described in the report found that students in traditional lecture courses were twice as likely to leave engineering and three times as likely to drop out of college entirely compared with students taught using active learning techniques. In another cited study, students in a physics class that used active learning methods learned twice as much as those taught in a traditional class, as measured by test results.
Active learning has been implemented in large lectures and it has been shown that both domestic and International students perceive a wide array of benefits. In a recent study, broad improvements were shown in student engagement and understanding of unit material among international students.±[31]
Active learning approaches have also been shown to reduce the contact between students and faculty by two thirds, while maintaining learning outcomes that were at least as good, and in one case, significantly better, compared to those achieved in traditional classrooms. Additionally, students' perceptions of their learning were improved and active learning classrooms were demonstrated to lead to a more efficient use of physical space.[32]
See also[edit]
References[edit]
Citations[edit]
- ^Bloom, B. S., Krathwohl, D. R., & Masia, B. B. (1956). Taxonomy of educational objectives: The classification of educational goals. New York, NY: David McKay Company.
- ^Renkl, A., Atkinson, R. K., Maier, U. H., & Staley, R. (2002). From example study to problem solving: Smooth transitions help learning. Journal of Experimental Education, 70 (4), 293–315.
- ^Bonwell, Charles; Eison, James (1991). Active Learning: Creating Excitement in the Classroom(PDF). Information Analyses - ERIC Clearinghouse Products (071). p. 3. ISBN978-1-878380-08-1. ISSN0884-0040.
- ^Bean, John C. (2011). Engaging Ideas: The Professor's Guide to Integrating Writing, Critical Thinking and Active Learning in the Classroom (2 ed.). John Wiley & Sons. p. 384. ISBN978-1-118-06233-3.
- ^Barnes, Douglas (1989). Active Learning. Leeds University TVEI Support Project, 1989. p. 19. ISBN978-1-872364-00-1.
- ^Kyriacou, Chris (1992). 'Active Learning in Secondary School Mathematics'. British Educational Research Journal. 18 (3): 309–318. doi:10.1080/0141192920180308. JSTOR1500835.
- ^Grabinger, Scott; Dunlap, Joanna (1995). 'Rich environments for active learning: a definition'. ALT-J. 3 (2): 5–34. doi:10.1080/0968776950030202.
- ^Panitz, Theodore (December 1999). COLLABORATIVE VERSUS COOPERATIVE LEARNING- A COMPARISON OF THE TWO CONCEPTS WHICH WILL HELP US UNDERSTAND THE UNDERLYING NATURE OF INTERACTIVE LEARNING(PDF). Eric. Retrieved 25 September 2015.
- ^Anthony, Glenda (1996). 'Active Learning in a Constructivist Framework'. Educational Studies in Mathematics. 31 (4): 349–369. doi:10.1007/BF00369153. JSTOR3482969.
- ^Rusbult, Craig. 'Constructivism as a Theory of Active Learning'. Retrieved 25 September 2015.
- ^ abcKerrey, Bob (2017-10-06). Kosslyn, Stephen M.; Nelson, Ben (eds.). Building the Intentional University: Minerva and the Future of Higher Education. The MIT Press. ISBN9780262037150.
- ^Brant, G., Hooper, E., & Sugrue, B. (1991). Which comes first: The simulation or the lecture? Journal of Educational Computing Research, 7(4), 469-481.
- ^Schwartz, D. L., & Bransford, J. D. (1998). A time for telling. Cognition and instruction, 16(4), 475-5223.
- ^Kapur, M., & Bielaczyc, K. (2011). Classroom-based experiments in productive failure. In Proceedings of the 33rd annual conference of the cognitive science society (pp. 2812-2817).
- ^Kapur, M. (2010). Productive failure in mathematical problem solving. Instructional Science, 38(6), 523-550.
- ^Kapur, M. (2008). Productive failure. Cognition and Instruction, 26(3), 379-424.
- ^Kapur, M. (2012). Productive failure in learning the concept of variance. Instructional Science, 40(4), 651-672.
- ^Kapur, M., & Bielaczyc, K. (2012). Designing for productive failure. Journal of the Learning Sciences, 21(1), 45-83.
- ^Westermann, K., & Rummel, N. (2012). Delaying instruction: evidence from a study in a university relearning setting. Instructional Science, 40(4), 673-689.
- ^McKeachie, W.J., Svinicki,M. (2006). Teaching Tips: Strategies, Research, and Theory for College and University Teachers. Belmont, CA. Wadsworth.
- ^Weimer, Maryellen. '10 benefits of getting students to participate in classroom discussions'. Faculty Focus. Faculty Focus. Retrieved 11 March 2015.
- ^Robertson, Kristina (2006). 'Increase Student Interaction with 'Think-Pair-Shares' and 'Circle Chats''. colorincolorado.org. Retrieved 5 March 2015.
- ^Harmann, Kerstin (2012). 'Assessing Student Perceptions of the benefits of discussions in small-group, large-class, and online learning contexts'. College Teaching. 60 (2): 65–75. doi:10.1080/87567555.2011.633407. Retrieved 10 March 2015.
- ^Karahocaa; et al. (2010). 'Computer assisted active learning system development for critical thinking in history of civilization'. Cypriot Journal of Educational Sciences.
- ^Freeman, S. et al. (2014). Active learning increases student performance in science, engineering, and mathematics. Proceedings of the National Academy of Scientists, 111(23), 8410–8415. https://dx.doi.org/10.1073/pnas.1319030111
- ^Hake, R. R. (1998). Interactive-engagement versus traditional methods: A six-thousand-student survey of mechanics test data for introductory physics courses. American Journal of Physics, 66, 64.
- ^Hoellwarth, C., & Moelter, M. J. (2011). The implications of a robust curriculum in introductory mechanics. American Journal of Physics, 79, 540.
- ^Prince, M. (2004). Does active learning work? A review of the research. Journal of engineering education, 93(3), 223-231.
- ^Michael, J. (2006). Where's the evidence that active learning works?. Advances in Physiology Education, 30(4), 159-167.
- ^President's Council of Advisors on Science and Technology. (2012). Engage to excel: Producing on million additional college graduates with degrees in science, technology, engineering, and mathematics. Retrieved from whitehouse.gov
- ^Marrone, Mauricio; Taylor, Murray; Hammerle, Mara (2018). 'Do International Students Appreciate Active Learning in Lectures?'. Australasian Journal of Information Systems. 22. doi:10.3127/ajis.v22i0.1334.
- ^Baepler, Paul; Walker, J.D.; Driessen, Michelle (2014). 'It's not about seat time: Blending, flipping, and efficiency in active learning classrooms'. Computers & Education. 78: 227–236. doi:10.1016/j.compedu.2014.06.006.
Works cited[edit]
- Bonwell, C.; Eison, J. (1991). Active Learning: Creating Excitement in the Classroom AEHE-ERIC Higher Education Report No. 1. Washington, D.C.: Jossey-Bass. ISBN978-1-878380-08-1.
- Chickering, Arthur W.; Zelda F. Gamson (March 1987). 'Seven Principles for Good Practice'. AAHE Bulletin. 39 (7): 3–7.
- McKinney, K. (2010). 'Active Learning. Illinois State University. Center for Teaching, Learning & Technology'. Archived from the original on 2011-09-11.
- Cranton, P. (2012). Planning instruction for adult learners (3rd ed.).
Toronto: Wall & Emerson.
- Brookfield, S. D. (2005). Discussion as the way of teaching: Tools and techniques for democratic classrooms (2nd ed.). San Francisco: Jossey-Bass.
- Bens, I. (2005). Understanding participation. In Facilitating with ease! Core skills for facilitators, team leaders and members, managers, consultants, and trainers (2nd ed., pp. 69–77). San Francisco: Jossey Bass.
- Radhakrishna R., Ewing J., and Chikthimmah N. (2012) NACTA Journal. 56.3
Further references[edit]
- Martyn, Margie (2007). 'Clickers in the Classroom: An Active Learning Approach'. EDUCAUSE Quarterly (EQ). 30 (2).
- Prince, M. (2004). Does Active Learning Work? A Review of the Research. Journal of Engineering Education, 93(3), 223-232.
External links[edit]
- Educational psychology in classroom settings. A developing open-source Wikibook related to learning as discussed in this article.
- Platform for Active Learning (University of Hull). Includes bank of examples.
Retrieved from 'https://en.wikipedia.org/w/index.php?title=Active_learning&oldid=889737613'
Machine learning and data mining |
---|
|
|
|
|
Deep learning (also known as deep structured learning or hierarchical learning) is part of a broader family of machine learning methods based on artificial neural networks. Learning can be supervised, semi-supervised or unsupervised.[1][2][3]
Deep learning architectures such as deep neural networks, deep belief networks, recurrent neural networks and convolutional neural networks have been applied to fields including computer vision, speech recognition, natural language processing, audio recognition, social network filtering, machine translation, bioinformatics, drug design, medical image analysis, material inspection and board game programs, where they have produced results comparable to and in some cases superior to human experts.[4][5][6]
Neural networks were inspired by information processing and distributed communication nodes in biological systems. ANNs have various differences from biological brains. Specifically, neural networks tend to be static and symbolic, while the biological brain of most living organisms is dynamic (plastic) and analog.[7][8][9]
- 4History
- 5Neural networks
- 5.2Deep neural networks
- 6Applications
- 9Criticism and comment
Definition[edit]
Deep learning is a class of machine learningalgorithms that:[10](pp199–200) use multiple layers to progressively extract higher level features from raw input. For example, in image processing, lower layers may identify edges, while higher layer may identify human-meaningful items such as digits/letters or faces.
Overview[edit]
Most modern deep learning models are based on an artificial neural networks, specifically, Convolutional Neural Networks (CNN)s, although they can also include propositional formulas or latent variables organized layer-wise in deep generative models such as the nodes in deep belief networks and deep Boltzmann machines.[11]
In deep learning, each level learns to transform its input data into a slightly more abstract and composite representation. In an image recognition application, the raw input may be a matrix of pixels; the first representational layer may abstract the pixels and encode edges; the second layer may compose and encode arrangements of edges; the third layer may encode a nose and eyes; and the fourth layer may recognize that the image contains a face. Importantly, a deep learning process can learn which features to optimally place in which level on its own. (Of course, this does not completely obviate the need for hand-tuning; for example, varying numbers of layers and layer sizes can provide different degrees of abstraction.)[1][12]
The 'deep' in 'deep learning' refers to the number of layers through which the data is transformed. More precisely, deep learning systems have a substantial credit assignment path (CAP) depth. The CAP is the chain of transformations from input to output. CAPs describe potentially causal connections between input and output. For a feedforward neural network, the depth of the CAPs is that of the network and is the number of hidden layers plus one (as the output layer is also parameterized). For recurrent neural networks, in which a signal may propagate through a layer more than once, the CAP depth is potentially unlimited.[2] No universally agreed upon threshold of depth divides shallow learning from deep learning, but most researchers agree that deep learning involves CAP depth > 2. CAP of depth 2 has been shown to be a universal approximator in the sense that it can emulate any function.[citation needed] Beyond that more layers do not add to the function approximator ability of the network. Deep models (CAP > 2) are able to extract better features than shallow models and hence, extra layers help in learning features.
Deep learning architectures are often constructed with a greedy layer-by-layer method.[clarification needed][further explanation needed][citation needed] Deep learning helps to disentangle these abstractions and pick out which features improve performance.[1]
For supervised learning tasks, deep learning methods obviate feature engineering, by translating the data into compact intermediate representations akin to principal components, and derive layered structures that remove redundancy in representation.
Deep learning algorithms can be applied to unsupervised learning tasks. This is an important benefit because unlabeled data are more abundant than labeled data. Examples of deep structures that can be trained in an unsupervised manner are neural history compressors[13] and deep belief networks.[1][14]
Interpretations[edit]
Deep neural networks are generally interpreted in terms of the universal approximation theorem[15][16][17][18][19][20] or probabilistic inference.[10][11][1][2][14][21][22]
The classic universal approximation theorem concerns the capacity of feedforward neural networks with a single hidden layer of finite size to approximate continuous functions.[15][16][17][18][19] In 1989, the first proof was published by George Cybenko for sigmoid activation functions[16] and was generalised to feed-forward multi-layer architectures in 1991 by Kurt Hornik.[17]
The universal approximation theorem for deep neural networks concerns the capacity of networks with bounded width but the depth is allowed to grow. Lu et al.[20] proved that if the width of a deep neural network with ReLU activation is strictly larger than the input dimension, then the network can approximate any Lebesgue integrable function; If the width is smaller or equal to the input dimension, then deep neural network is not a universal approximator.
The probabilistic interpretation[21] derives from the field of machine learning. It features inference,[10][11][1][2][14][21] as well as the optimization concepts of training and testing, related to fitting and generalization, respectively. More specifically, the probabilistic interpretation considers the activation nonlinearity as a cumulative distribution function.[21] The probabilistic interpretation led to the introduction of dropout as regularizer in neural networks.[23] The probabilistic interpretation was introduced by researchers including Hopfield, Widrow and Narendra and popularized in surveys such as the one by Bishop.[24]
History[edit]
The term Deep Learning was introduced to the machine learning community by Rina Dechter in 1986,[25][13] and to artificial neural networks by Igor Aizenberg and colleagues in 2000, in the context of Boolean threshold neurons.[26][27]
The first general, working learning algorithm for supervised, deep, feedforward, multilayer perceptrons was published by Alexey Ivakhnenko and Lapa in 1965.[28] A 1971 paper described a deep network with 8 layers trained by the group method of data handling algorithm.[29]
Other deep learning working architectures, specifically those built for computer vision, began with the Neocognitron introduced by Kunihiko Fukushima in 1980.[30] In 1989, Yann LeCun et al. applied the standard backpropagation algorithm, which had been around as the reverse mode of automatic differentiation since 1970,[31][32][33][34] to a deep neural network with the purpose of recognizing handwritten ZIP codes on mail. While the algorithm worked, training required 3 days.[35]
By 1991 such systems were used for recognizing isolated 2-D hand-written digits, while recognizing 3-D objects was done by matching 2-D images with a handcrafted 3-D object model. Weng et al. suggested that a human brain does not use a monolithic 3-D object model and in 1992 they published Cresceptron,[36][37][38] a method for performing 3-D object recognition in cluttered scenes. Because it directly used natural images, Cresceptron started the beginning of general-purpose visual learning for natural 3D worlds. Cresceptron is a cascade of layers similar to Neocognitron. But while Neocognitron required a human programmer to hand-merge features, Cresceptron learned an open number of features in each layer without supervision, where each feature is represented by a convolution kernel. Cresceptron segmented each learned object from a cluttered scene through back-analysis through the network. Max pooling, now often adopted by deep neural networks (e.g. ImageNet tests), was first used in Cresceptron to reduce the position resolution by a factor of (2x2) to 1 through the cascade for better generalization.
In 1994, André de Carvalho, together with Mike Fairhurst and David Bisset, published experimental results of a multi-layer boolean neural network, also known as a weightless neural network, composed of a 3-layers self-organising feature extraction neural network module (SOFT) followed by a multi-layer classification neural network module (GSN), which were independently trained. Each layer in the feature extraction module extracted features with growing complexity regarding the previous layer.[39]
In 1995, Brendan Frey demonstrated that it was possible to train (over two days) a network containing six fully connected layers and several hundred hidden units using the wake-sleep algorithm, co-developed with Peter Dayan and Hinton.[40] Many factors contribute to the slow speed, including the vanishing gradient problem analyzed in 1991 by Sepp Hochreiter.[41][42]
Simpler models that use task-specific handcrafted features such as Gabor filters and support vector machines (SVMs) were a popular choice in the 1990s and 2000s, because of artificial neural network's (ANN) computational cost and a lack of understanding of how the brain wires its biological networks.
Both shallow and deep learning (e.g., recurrent nets) of ANNs have been explored for many years.[43][44][45] These methods never outperformed non-uniform internal-handcrafting Gaussian mixture model/Hidden Markov model (GMM-HMM) technology based on generative models of speech trained discriminatively.[46] Key difficulties have been analyzed, including gradient diminishing[41] and weak temporal correlation structure in neural predictive models.[47][48] Additional difficulties were the lack of training data and limited computing power.
Most speech recognition researchers moved away from neural nets to pursue generative modeling. An exception was at SRI International in the late 1990s. Funded by the US government's NSA and DARPA, SRI studied deep neural networks in speech and speaker recognition. Heck's speaker recognition team achieved the first significant success with deep neural networks in speech processing in the 1998 National Institute of Standards and Technology Speaker Recognition evaluation.[49] While SRI experienced success with deep neural networks in speaker recognition, they were unsuccessful in demonstrating similar success in speech recognition.The principle of elevating 'raw' features over hand-crafted optimization was first explored successfully in the architecture of deep autoencoder on the 'raw' spectrogram or linear filter-bank features in the late 1990s,[49] showing its superiority over the Mel-Cepstral features that contain stages of fixed transformation from spectrograms. The raw features of speech, waveforms, later produced excellent larger-scale results.[50]
Many aspects of speech recognition were taken over by a deep learning method called long short-term memory (LSTM), a recurrent neural network published by Hochreiter and Schmidhuber in 1997.[51] LSTM RNNs avoid the vanishing gradient problem and can learn 'Very Deep Learning' tasks[2] that require memories of events that happened thousands of discrete time steps before, which is important for speech. In 2003, LSTM started to become competitive with traditional speech recognizers on certain tasks.[52] Later it was combined with connectionist temporal classification (CTC)[53] in stacks of LSTM RNNs.[54] In 2015, Google's speech recognition reportedly experienced a dramatic performance jump of 49% through CTC-trained LSTM, which they made available through Google Voice Search.[55]
In 2006, publications by Geoff Hinton, Ruslan Salakhutdinov, Osindero and Teh[56][57][58] showed how a many-layered feedforward neural network could be effectively pre-trained one layer at a time, treating each layer in turn as an unsupervised restricted Boltzmann machine, then fine-tuning it using supervised backpropagation.[59] The papers referred to learning for deep belief nets.
Deep learning is part of state-of-the-art systems in various disciplines, particularly computer vision and automatic speech recognition (ASR). Results on commonly used evaluation sets such as TIMIT (ASR) and MNIST (image classification), as well as a range of large-vocabulary speech recognition tasks have steadily improved.[60][61][62]Convolutional neural networks (CNNs) were superseded for ASR by CTC[53] for LSTM.[51][55][63][64][65][66][67] but are more successful in computer vision.
The impact of deep learning in industry began in the early 2000s, when CNNs already processed an estimated 10% to 20% of all the checks written in the US, according to Yann LeCun.[68] Industrial applications of deep learning to large-scale speech recognition started around 2010.
The 2009 NIPS Workshop on Deep Learning for Speech Recognition[69] was motivated by the limitations of deep generative models of speech, and the possibility that given more capable hardware and large-scale data sets that deep neural nets (DNN) might become practical. It was believed that pre-training DNNs using generative models of deep belief nets (DBN) would overcome the main difficulties of neural nets.[70] However, it was discovered that replacing pre-training with large amounts of training data for straightforward backpropagation when using DNNs with large, context-dependent output layers produced error rates dramatically lower than then-state-of-the-art Gaussian mixture model (GMM)/Hidden Markov Model (HMM) and also than more-advanced generative model-based systems.[60][71] The nature of the recognition errors produced by the two types of systems was characteristically different,[72][69] offering technical insights into how to integrate deep learning into the existing highly efficient, run-time speech decoding system deployed by all major speech recognition systems.[10][73][74] Analysis around 2009-2010, contrasted the GMM (and other generative speech models) vs. DNN models, stimulated early industrial investment in deep learning for speech recognition,[72][69] eventually leading to pervasive and dominant use in that industry. That analysis was done with comparable performance (less than 1.5% in error rate) between discriminative DNNs and generative models.[60][72][70][75]
In 2010, researchers extended deep learning from TIMIT to large vocabulary speech recognition, by adopting large output layers of the DNN based on context-dependent HMM states constructed by decision trees.[76][77][78][73]
Advances in hardware enabled the renewed interest. In 2009, Nvidia was involved in what was called the “big bang” of deep learning, “as deep-learning neural networks were trained with Nvidia graphics processing units (GPUs).”[79] That year, Google Brain used Nvidia GPUs to create capable DNNs. While there, Andrew Ng determined that GPUs could increase the speed of deep-learning systems by about 100 times.[80] In particular, GPUs are well-suited for the matrix/vector math involved in machine learning.[81][82] GPUs speed up training algorithms by orders of magnitude, reducing running times from weeks to days.[83][84] Specialized hardware and algorithm optimizations can be used for efficient processing.[85]
Deep learning revolution[edit]
How deep learning is a subset of machine learning and how machine learning is a subset of artificial intelligence (AI).
In 2012, a team led by Dahl won the 'Merck Molecular Activity Challenge' using multi-task deep neural networks to predict the biomolecular target of one drug.[86][87] In 2014, Hochreiter's group used deep learning to detect off-target and toxic effects of environmental chemicals in nutrients, household products and drugs and won the 'Tox21 Data Challenge' of NIH, FDA and NCATS.[88][89][90]
Significant additional impacts in image or object recognition were felt from 2011 to 2012. Although CNNs trained by backpropagation had been around for decades, and GPU implementations of NNs for years, including CNNs, fast implementations of CNNs with max-pooling on GPUs in the style of Ciresan and colleagues were needed to progress on computer vision.[81][82][35][91][2] In 2011, this approach achieved for the first time superhuman performance in a visual pattern recognition contest. Also in 2011, it won the ICDAR Chinese handwriting contest, and in May 2012, it won the ISBI image segmentation contest.[92] Until 2011, CNNs did not play a major role at computer vision conferences, but in June 2012, a paper by Ciresan et al. at the leading conference CVPR[4] showed how max-pooling CNNs on GPU can dramatically improve many vision benchmark records. In October 2012, a similar system by Krizhevsky et al.[5] won the large-scale ImageNet competition by a significant margin over shallow machine learning methods. In November 2012, Ciresan et al.'s system also won the ICPR contest on analysis of large medical images for cancer detection, and in the following year also the MICCAI Grand Challenge on the same topic.[93] In 2013 and 2014, the error rate on the ImageNet task using deep learning was further reduced, following a similar trend in large-scale speech recognition. The Wolfram Image Identification project publicized these improvements.[94]
Image classification was then extended to the more challenging task of generating descriptions (captions) for images, often as a combination of CNNs and LSTMs.[95][96][97][98]
Some researchers assess that the October 2012 ImageNet victory anchored the start of a 'deep learning revolution' that has transformed the AI industry.[99]
In March 2019, Yoshua Bengio, Geoffrey Hinton and Yann LeCun were awarded the Turing Award for conceptual and engineering breakthroughs that have made deep neural networks a critical component of computing.
Neural networks[edit]
Artificial neural networks[edit]
Artificial neural networks (ANNs) or connectionist systems are computing systems inspired by the biological neural networks that constitute animal brains. Such systems learn (progressively improve their ability) to do tasks by considering examples, generally without task-specific programming. For example, in image recognition, they might learn to identify images that contain cats by analyzing example images that have been manually labeled as 'cat' or 'no cat' and using the analytic results to identify cats in other images. They have found most use in applications difficult to express with a traditional computer algorithm using rule-based programming.
An ANN is based on a collection of connected units called artificial neurons, (analogous to biological neurons in a biological brain). Each connection (synapse) between neurons can transmit a signal to another neuron. The receiving (postsynaptic) neuron can process the signal(s) and then signal downstream neurons connected to it. Neurons may have state, generally represented by real numbers, typically between 0 and 1. Neurons and synapses may also have a weight that varies as learning proceeds, which can increase or decrease the strength of the signal that it sends downstream.
Typically, neurons are organized in layers. Different layers may perform different kinds of transformations on their inputs. Signals travel from the first (input), to the last (output) layer, possibly after traversing the layers multiple times.
The original goal of the neural network approach was to solve problems in the same way that a human brain would. Over time, attention focused on matching specific mental abilities, leading to deviations from biology such as backpropagation, or passing information in the reverse direction and adjusting the network to reflect that information.
Neural networks have been used on a variety of tasks, including computer vision, speech recognition, machine translation, social network filtering, playing board and video games and medical diagnosis.
As of 2017, neural networks typically have a few thousand to a few million units and millions of connections. Despite this number being several order of magnitude less than the number of neurons on a human brain, these networks can perform many tasks at a level beyond that of humans (e.g., recognizing faces, playing 'Go'[100] ).
Deep neural networks[edit]
A deep neural network (DNN) is an artificial neural network (ANN) with multiple layers between the input and output layers.[11][2] The DNN finds the correct mathematical manipulation to turn the input into the output, whether it be a linear relationship or a non-linear relationship. The network moves through the layers calculating the probability of each output. For example, a DNN that is trained to recognize dog breeds will go over the given image and calculate the probability that the dog in the image is a certain breed. The user can review the results and select which probabilities the network should display (above a certain threshold, etc.) and return the proposed label. Each mathematical manipulation as such is considered a layer, and complex DNN have many layers, hence the name 'deep' networks.
DNNs can model complex non-linear relationships. DNN architectures generate compositional models where the object is expressed as a layered composition of primitives.[101] The extra layers enable composition of features from lower layers, potentially modeling complex data with fewer units than a similarly performing shallow network.[11]
Deep architectures include many variants of a few basic approaches. Each architecture has found success in specific domains. It is not always possible to compare the performance of multiple architectures, unless they have been evaluated on the same data sets.
DNNs are typically feedforward networks in which data flows from the input layer to the output layer without looping back. At first, the DNN creates a map of virtual neurons and assigns random numerical values, or 'weights', to connections between them. The weights and inputs are multiplied and return an output between 0 and 1. If the network didn’t accurately recognize a particular pattern, an algorithm would adjust the weights.[102] That way the algorithm can make certain parameters more influential, until it determines the correct mathematical manipulation to fully process the data.
Recurrent neural networks (RNNs), in which data can flow in any direction, are used for applications such as language modeling.[103][104][105][106][107] Long short-term memory is particularly effective for this use.[51][108]
Convolutional deep neural networks (CNNs) are used in computer vision.[109] CNNs also have been applied to acoustic modeling for automatic speech recognition (ASR).[67]
Challenges[edit]
As with ANNs, many issues can arise with naively trained DNNs. Two common issues are overfitting and computation time.
DNNs are prone to overfitting because of the added layers of abstraction, which allow them to model rare dependencies in the training data. Regularization methods such as Ivakhnenko's unit pruning[29] or weight decay (-regularization) or sparsity (-regularization) can be applied during training to combat overfitting.[110] Alternatively dropout regularization randomly omits units from the hidden layers during training. This helps to exclude rare dependencies.[111] Finally, data can be augmented via methods such as cropping and rotating such that smaller training sets can be increased in size to reduce the chances of overfitting.[112]
DNNs must consider many training parameters, such as the size (number of layers and number of units per layer), the learning rate, and initial weights. Sweeping through the parameter space for optimal parameters may not be feasible due to the cost in time and computational resources. Various tricks, such as batching (computing the gradient on several training examples at once rather than individual examples)[113] speed up computation. Large processing capabilities of many-core architectures (such as, GPUs or the Intel Xeon Phi) have produced significant speedups in training, because of the suitability of such processing architectures for the matrix and vector computations.[114][115]
Alternatively, engineers may look for other types of neural networks with more straightforward and convergent training algorithms. CMAC (cerebellar model articulation controller) is one such kind of neural network. It doesn't require learning rates or randomized initial weights for CMAC. The training process can be guaranteed to converge in one step with a new batch of data, and the computational complexity of the training algorithm is linear with respect to the number of neurons involved.[116][117]
Applications[edit]
Automatic speech recognition[edit]
Large-scale automatic speech recognition is the first and most convincing successful case of deep learning. LSTM RNNs can learn 'Very Deep Learning' tasks[2] that involve multi-second intervals containing speech events separated by thousands of discrete time steps, where one time step corresponds to about 10 ms. LSTM with forget gates[108] is competitive with traditional speech recognizers on certain tasks.[52]
The initial success in speech recognition was based on small-scale recognition tasks based on TIMIT. The data set contains 630 speakers from eight major dialects of American English, where each speaker reads 10 sentences.[118] Its small size lets many configurations be tried. More importantly, the TIMIT task concerns phone-sequence recognition, which, unlike word-sequence recognition, allows weak phone bigram language models. This lets the strength of the acoustic modeling aspects of speech recognition be more easily analyzed. The error rates listed below, including these early results and measured as percent phone error rates (PER), have been summarized since 1991.
Method | PER (%) |
---|---|
Randomly Initialized RNN[119] | 26.1 |
Bayesian Triphone GMM-HMM | 25.6 |
Hidden Trajectory (Generative) Model | 24.8 |
Monophone Randomly Initialized DNN | 23.4 |
Monophone DBN-DNN | 22.4 |
Triphone GMM-HMM with BMMI Training | 21.7 |
Monophone DBN-DNN on fbank | 20.7 |
Convolutional DNN[120] | 20.0 |
Convolutional DNN w. Heterogeneous Pooling | 18.7 |
Ensemble DNN/CNN/RNN[121] | 18.3 |
Bidirectional LSTM | 17.9 |
Hierarchical Convolutional Deep Maxout Network[122] | 16.5 |
The debut of DNNs for speaker recognition in the late 1990s and speech recognition around 2009-2011 and of LSTM around 2003-2007, accelerated progress in eight major areas:[10][75][73]
- Scale-up/out and acclerated DNN training and decoding
- Sequence discriminative training
- Feature processing by deep models with solid understanding of the underlying mechanisms
- Adaptation of DNNs and related deep models
- Multi-task and transfer learning by DNNs and related deep models
- CNNs and how to design them to best exploit domain knowledge of speech
- RNN and its rich LSTM variants
- Other types of deep models including tensor-based models and integrated deep generative/discriminative models.
All major commercial speech recognition systems (e.g., Microsoft Cortana, Xbox, Skype Translator, Amazon Alexa, Google Now, Apple Siri, Baidu and iFlyTek voice search, and a range of Nuance speech products, etc.) are based on deep learning.[10][123][124][125]
Image recognition[edit]
A common evaluation set for image classification is the MNIST database data set. MNIST is composed of handwritten digits and includes 60,000 training examples and 10,000 test examples. As with TIMIT, its small size lets users test multiple configurations. A comprehensive list of results on this set is available.[126]
Deep learning-based image recognition has become 'superhuman', producing more accurate results than human contestants. This first occurred in 2011.[127]
Deep learning-trained vehicles now interpret 360° camera views.[128] Another example is Facial Dysmorphology Novel Analysis (FDNA) used to analyze cases of human malformation connected to a large database of genetic syndromes.
Visual art processing[edit]
Closely related to the progress that has been made in image recognition is the increasing application of deep learning techniques to various visual art tasks. DNNs have proven themselves capable, for example, of a) identifying the style period of a given painting, b) Neural Style Transfer - capturing the style of a given artwork and applying it in a visually pleasing manner to an arbitrary photograph or video, and c) generating striking imagery based on random visual input fields.[129][130]
Natural language processing[edit]
Neural networks have been used for implementing language models since the early 2000s.[103][131] LSTM helped to improve machine translation and language modeling.[104][105][106]
Other key techniques in this field are negative sampling[132] and word embedding. Word embedding, such as word2vec, can be thought of as a representational layer in a deep learning architecture that transforms an atomic word into a positional representation of the word relative to other words in the dataset; the position is represented as a point in a vector space. Using word embedding as an RNN input layer allows the network to parse sentences and phrases using an effective compositional vector grammar. A compositional vector grammar can be thought of as probabilistic context free grammar (PCFG) implemented by an RNN.[133] Recursive auto-encoders built atop word embeddings can assess sentence similarity and detect paraphrasing.[133] Deep neural architectures provide the best results for constituency parsing,[134]sentiment analysis,[135] information retrieval,[136][137] spoken language understanding,[138] machine translation,[104][139] contextual entity linking,[139] writing style recognition,[140] Text classification and others.[141]
Recent developments generalize word embedding to sentence embedding.
Google Translate (GT) uses a large end-to-end long short-term memory network.[142][143][144][145][146][147]Google Neural Machine Translation (GNMT) uses an example-based machine translation method in which the system 'learns from millions of examples.'[143] It translates 'whole sentences at a time, rather than pieces. Google Translate supports over one hundred languages.[143] The network encodes the 'semantics of the sentence rather than simply memorizing phrase-to-phrase translations'.[143][148] GT uses English as an intermediate between most language pairs.[148]
Drug discovery and toxicology[edit]
A large percentage of candidate drugs fail to win regulatory approval. These failures are caused by insufficient efficacy (on-target effect), undesired interactions (off-target effects), or unanticipated toxic effects.[149][150] Research has explored use of deep learning to predict the biomolecular targets,[86][87]off-targets, and toxic effects of environmental chemicals in nutrients, household products and drugs.[88][89][90]
AtomNet is a deep learning system for structure-based rational drug design.[151] AtomNet was used to predict novel candidate biomolecules for disease targets such as the Ebola virus[152] and multiple sclerosis.[153][154]
Customer relationship management[edit]
Deep reinforcement learning has been used to approximate the value of possible direct marketing actions, defined in terms of RFM variables. The estimated value function was shown to have a natural interpretation as customer lifetime value.[155]
Recommendation systems[edit]
Recommendation systems have used deep learning to extract meaningful features for a latent factor model for content-based music recommendations.[156] Multiview deep learning has been applied for learning user preferences from multiple domains.[157] The model uses a hybrid collaborative and content-based approach and enhances recommendations in multiple tasks.
Bioinformatics[edit]
An autoencoder ANN was used in bioinformatics, to predict gene ontology annotations and gene-function relationships.[158]
In medical informatics, deep learning was used to predict sleep quality based on data from wearables[159] and predictions of health complications from electronic health record data.[160] Deep learning has also showed efficacy in healthcare.[161]
Medical Image Analysis[edit]
Deep learning has been shown to produce competitive results in medical application such as cancer cell classification, lesion detection, organ segmentation and image enhancement[162][163]
Mobile advertising[edit]
Finding the appropriate mobile audience for mobile advertising is always challenging, since many data points must be considered and assimilated before a target segment can be created and used in ad serving by any ad server.[164] Deep learning has been used to interpret large, many-dimensioned advertising datasets. Many data points are collected during the request/serve/click internet advertising cycle. This information can form the basis of machine learning to improve ad selection.
Image restoration[edit]
Deep learning has been successfully applied to inverse problems such as denoising, super-resolution, inpainting, and film colorization. These applications include learning methods such as 'Shrinkage Fields for Effective Image Restoration'[165] which trains on an image dataset, and Deep Image Prior, which trains on the image that needs restoration.
Financial fraud detection[edit]
Deep learning is being successfully applied to financial fraud detection and anti-money laundering. 'Deep anti-money laundering detection system can spot and recognize relationships and similarities between data and, further down the road, learn to detect anomalies or classify and predict specific events'. The solution leverages both supervised learning techniques, such as the classification of suspicious transactions, and unsupervised learning, e.g. anomaly detection.[166]
Military[edit]
The United States Department of Defense applied deep learning to train robots in new tasks through observation.[167]
Relation to human cognitive and brain development[edit]
Deep learning is closely related to a class of theories of brain development (specifically, neocortical development) proposed by cognitive neuroscientists in the early 1990s.[168][169][170][171] These developmental theories were instantiated in computational models, making them predecessors of deep learning systems. These developmental models share the property that various proposed learning dynamics in the brain (e.g., a wave of nerve growth factor) support the self-organization somewhat analogous to the neural networks utilized in deep learning models. Like the neocortex, neural networks employ a hierarchy of layered filters in which each layer considers information from a prior layer (or the operating environment), and then passes its output (and possibly the original input), to other layers. This process yields a self-organizing stack of transducers, well-tuned to their operating environment. A 1995 description stated, '..the infant's brain seems to organize itself under the influence of waves of so-called trophic-factors .. different regions of the brain become connected sequentially, with one layer of tissue maturing before another and so on until the whole brain is mature.'[172]
A variety of approaches have been used to investigate the plausibility of deep learning models from a neurobiological perspective. On the one hand, several variants of the backpropagation algorithm have been proposed in order to increase its processing realism.[173][174] Other researchers have argued that unsupervised forms of deep learning, such as those based on hierarchical generative models and deep belief networks, may be closer to biological reality.[175][176] In this respect, generative neural network models have been related to neurobiological evidence about sampling-based processing in the cerebral cortex.[177]
Although a systematic comparison between the human brain organization and the neuronal encoding in deep networks has not yet been established, several analogies have been reported. For example, the computations performed by deep learning units could be similar to those of actual neurons[178][179] and neural populations.[180] Similarly, the representations developed by deep learning models are similar to those measured in the primate visual system[181] both at the single-unit[182] and at the population[183] levels.
Commercial activity[edit]
Many organizations employ deep learning for particular applications. Facebook's AI lab performs tasks such as automatically tagging uploaded pictures with the names of the people in them.[184]
Google's DeepMind Technologies developed a system capable of learning how to play Atari video games using only pixels as data input. In 2015 they demonstrated their AlphaGo system, which learned the game of Go well enough to beat a professional Go player.[185][186][187] Google Translate uses an LSTM to translate between more than 100 languages.
In 2015, Blippar demonstrated a mobile augmented reality application that uses deep learning to recognize objects in real time.[188]
As of 2008,[189] researchers at The University of Texas at Austin (UT) developed a machine learning framework called Training an Agent Manually via Evaluative Reinforcement, or TAMER, which proposed new methods for robots or computer programs to learn how to perform tasks by interacting with a human instructor.[167]
First developed as TAMER, a new algorithm called Deep TAMER was later introduced in 2018 during a collaboration between U.S. Army Research Laboratory (ARL) and UT researchers. Deep TAMER used deep learning to provide a robot the ability to learn new tasks through observation.[167]
Adobe flash update check. Using Deep TAMER, a robot learned a task with a human trainer, watching video streams or observing a human perform a task in-person. The robot later practiced the task with the help of some coaching from the trainer, who provided feedback such as “good job” and “bad job.”[190]
Criticism and comment[edit]
Deep learning has attracted both criticism and comment, in some cases from outside the field of computer science.
Theory[edit]
A main criticism concerns the lack of theory surrounding some methods.[191] Learning in the most common deep architectures is implemented using well-understood gradient descent. However, the theory surrounding other algorithms, such as contrastive divergence is less clear.[citation needed] (e.g., Does it converge? If so, how fast? What is it approximating?) Deep learning methods are often looked at as a black box, with most confirmations done empirically, rather than theoretically.[192]
Others point out that deep learning should be looked at as a step towards realizing strong AI, not as an all-encompassing solution. Despite the power of deep learning methods, they still lack much of the functionality needed for realizing this goal entirely. Research psychologist Gary Marcus noted:
'Realistically, deep learning is only part of the larger challenge of building intelligent machines. Such techniques lack ways of representing causal relationships (..) have no obvious ways of performing logical inferences, and they are also still a long way from integrating abstract knowledge, such as information about what objects are, what they are for, and how they are typically used. The most powerful A.I. systems, like Watson (..) use techniques like deep learning as just one element in a very complicated ensemble of techniques, ranging from the statistical technique of Bayesian inference to deductive reasoning.'[193]
As an alternative to this emphasis on the limits of deep learning, one author speculated that it might be possible to train a machine vision stack to perform the sophisticated task of discriminating between 'old master' and amateur figure drawings, and hypothesized that such a sensitivity might represent the rudiments of a non-trivial machine empathy.[194] This same author proposed that this would be in line with anthropology, which identifies a concern with aesthetics as a key element of behavioral modernity.[195]
In further reference to the idea that artistic sensitivity might inhere within relatively low levels of the cognitive hierarchy, a published series of graphic representations of the internal states of deep (20-30 layers) neural networks attempting to discern within essentially random data the images on which they were trained[196] demonstrate a visual appeal: the original research notice received well over 1,000 comments, and was the subject of what was for a time the most frequently accessed article on The Guardian's[197] web site.
Errors[edit]
Some deep learning architectures display problematic behaviors,[198] such as confidently classifying unrecognizable images as belonging to a familiar category of ordinary images[199] and misclassifying minuscule perturbations of correctly classified images.[200]Goertzel hypothesized that these behaviors are due to limitations in their internal representations and that these limitations would inhibit integration into heterogeneous multi-component artificial general intelligence (AGI) architectures.[198] These issues may possibly be addressed by deep learning architectures that internally form states homologous to image-grammar[201] decompositions of observed entities and events.[198]Learning a grammar (visual or linguistic) from training data would be equivalent to restricting the system to commonsense reasoning that operates on concepts in terms of grammatical production rules and is a basic goal of both human language acquisition[202] and artificial intelligence (AI).[203]
Cyberthreat[edit]
As deep learning moves from the lab into the world, research and experience shows that artificial neural networks are vulnerable to hacks and deception. By identifying patterns that these systems use to function, attackers can modify inputs to ANNs in such a way that the ANN finds a match that human observers would not recognize. For example, an attacker can make subtle changes to an image such that the ANN finds a match even though the image looks to a human nothing like the search target. Such a manipulation is termed an “adversarial attack.” In 2016 researchers used one ANN to doctor images in trial and error fashion, identify another's focal points and thereby generate images that deceived it. The modified images looked no different to human eyes. Another group showed that printouts of doctored images then photographed successfully tricked an image classification system.[204] One defense is reverse image search, in which a possible fake image is submitted to a site such as TinEye that can then find other instances of it. A refinement is to search using only parts of the image, to identify images from which that piece may have been taken.[205]
Another group showed that certain psychedelic spectacles could fool a facial recognition system into thinking ordinary people were celebrities, potentially allowing one person to impersonate another. In 2017 researchers added stickers to stop signs and caused an ANN to misclassify them.[204]
ANNs can however be further trained to detect attempts at deception, potentially leading attackers and defenders into an arms race similar to the kind that already defines the malware defense industry. ANNs have been trained to defeat ANN-based anti-malware software by repeatedly attacking a defense with malware that was continually altered by a genetic algorithm until it tricked the anti-malware while retaining its ability to damage the target.[204]
Que Es El Aprendizaje
Another group demonstrated that certain sounds could make the Google Now voice command system open a particular web address that would download malware.[204]
In “data poisoning”, false data is continually smuggled into a machine learning system’s training set to prevent it from achieving mastery.[204]
See also[edit]
References[edit]
- ^ abcdefBengio, Y.; Courville, A.; Vincent, P. (2013). 'Representation Learning: A Review and New Perspectives'. IEEE Transactions on Pattern Analysis and Machine Intelligence. 35 (8): 1798–1828. arXiv:1206.5538. doi:10.1109/tpami.2013.50. PMID23787338.
- ^ abcdefghSchmidhuber, J. (2015). 'Deep Learning in Neural Networks: An Overview'. Neural Networks. 61: 85–117. arXiv:1404.7828. doi:10.1016/j.neunet.2014.09.003. PMID25462637.
- ^Bengio, Yoshua; LeCun, Yann; Hinton, Geoffrey (2015). 'Deep Learning'. Nature. 521 (7553): 436–444. Bibcode:2015Natur.521.436L. doi:10.1038/nature14539. PMID26017442.
- ^ abCiresan, Dan; Meier, U.; Schmidhuber, J. (June 2012). 'Multi-column deep neural networks for image classification'. 2012 IEEE Conference on Computer Vision and Pattern Recognition: 3642–3649. arXiv:1202.2745. doi:10.1109/cvpr.2012.6248110. ISBN978-1-4673-1228-8.
- ^ abKrizhevsky, Alex; Sutskever, Ilya; Hinton, Geoffry (2012). 'ImageNet Classification with Deep Convolutional Neural Networks'(PDF). NIPS 2012: Neural Information Processing Systems, Lake Tahoe, Nevada.
- ^'Google's AlphaGo AI wins three-match series against the world's best Go player'. TechCrunch. 25 May 2017.
- ^Marblestone, Adam H.; Wayne, Greg; Kording, Konrad P. (2016). 'Toward an Integration of Deep Learning and Neuroscience'. Frontiers in Computational Neuroscience. 10: 94. doi:10.3389/fncom.2016.00094. PMC5021692. PMID27683554.
- ^Olshausen, B. A. (1996). 'Emergence of simple-cell receptive field properties by learning a sparse code for natural images'. Nature. 381 (6583): 607–609. Bibcode:1996Natur.381.607O. doi:10.1038/381607a0. PMID8637596.
- ^Bengio, Yoshua; Lee, Dong-Hyun; Bornschein, Jorg; Mesnard, Thomas; Lin, Zhouhan (2015-02-13). 'Towards Biologically Plausible Deep Learning'. arXiv:1502.04156 [cs.LG].
- ^ abcdefDeng, L.; Yu, D. (2014). 'Deep Learning: Methods and Applications'(PDF). Foundations and Trends in Signal Processing. 7 (3–4): 1–199. doi:10.1561/2000000039.
- ^ abcdeBengio, Yoshua (2009). 'Learning Deep Architectures for AI'(PDF). Foundations and Trends in Machine Learning. 2 (1): 1–127. CiteSeerX10.1.1.701.9550. doi:10.1561/2200000006.
- ^LeCun, Yann; Bengio, Yoshua; Hinton, Geoffrey (28 May 2015). 'Deep learning'. Nature. 521 (7553): 436–444. Bibcode:2015Natur.521.436L. doi:10.1038/nature14539. PMID26017442.
- ^ abJürgen Schmidhuber (2015). Deep Learning. Scholarpedia, 10(11):32832. Online
- ^ abcHinton, G.E. (2009). 'Deep belief networks'. Scholarpedia. 4 (5): 5947. Bibcode:2009SchpJ..4.5947H. doi:10.4249/scholarpedia.5947.
- ^ abBalázs Csanád Csáji (2001). Approximation with Artificial Neural Networks; Faculty of Sciences; Eötvös Loránd University, Hungary
- ^ abcCybenko (1989). 'Approximations by superpositions of sigmoidal functions'(PDF). Mathematics of Control, Signals, and Systems. 2 (4): 303–314. doi:10.1007/bf02551274. Archived from the original(PDF) on 2015-10-10.
- ^ abcHornik, Kurt (1991). 'Approximation Capabilities of Multilayer Feedforward Networks'. Neural Networks. 4 (2): 251–257. doi:10.1016/0893-6080(91)90009-t.
- ^ abHaykin, Simon S. (1999). Neural Networks: A Comprehensive Foundation. Prentice Hall. ISBN978-0-13-273350-2.
- ^ abHassoun, Mohamad H. (1995). Fundamentals of Artificial Neural Networks. MIT Press. p. 48. ISBN978-0-262-08239-6.
- ^ abLu, Z., Pu, H., Wang, F., Hu, Z., & Wang, L. (2017). The Expressive Power of Neural Networks: A View from the Width. Neural Information Processing Systems, 6231-6239.
- ^ abcdMurphy, Kevin P. (24 August 2012). Machine Learning: A Probabilistic Perspective. MIT Press. ISBN978-0-262-01802-9.
- ^Patel, Ankit; Nguyen, Tan; Baraniuk, Richard (2016). 'A Probabilistic Framework for Deep Learning'(PDF). Advances in Neural Information Processing Systems.
- ^Hinton, G. E.; Srivastava, N.; Krizhevsky, A.; Sutskever, I.; Salakhutdinov, R.R. (2012). 'Improving neural networks by preventing co-adaptation of feature detectors'. arXiv:1207.0580 [math.LG].
- ^Bishop, Christopher M. (2006). Pattern Recognition and Machine Learning(PDF). Springer. ISBN978-0-387-31073-2.
- ^Rina Dechter (1986). Learning while searching in constraint-satisfaction problems. University of California, Computer Science Department, Cognitive Systems Laboratory.Online
- ^Igor Aizenberg, Naum N. Aizenberg, Joos P.L. Vandewalle (2000). Multi-Valued and Universal Binary Neurons: Theory, Learning and Applications. Springer Science & Business Media.
- ^Co-evolving recurrent neurons learn deep memory POMDPs. Proc. GECCO, Washington, D. C., pp. 1795-1802, ACM Press, New York, NY, USA, 2005.
- ^Ivakhnenko, A. G. (1973). Cybernetic Predicting Devices. CCM Information Corporation.
- ^ abIvakhnenko, Alexey (1971). 'Polynomial theory of complex systems'. IEEE Transactions on Systems, Man and Cybernetics. 1 (4): 364–378. doi:10.1109/TSMC.1971.4308320.
- ^Fukushima, K. (1980). 'Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position'. Biol. Cybern. 36 (4): 193–202. doi:10.1007/bf00344251. PMID7370364.
- ^Seppo Linnainmaa (1970). The representation of the cumulative rounding error of an algorithm as a Taylor expansion of the local rounding errors. Master's Thesis (in Finnish), Univ. Helsinki, 6-7.
- ^Griewank, Andreas (2012). 'Who Invented the Reverse Mode of Differentiation?'(PDF). Documenta Matematica (Extra Volume ISMP): 389–400.
- ^Werbos, P. (1974). 'Beyond Regression: New Tools for Prediction and Analysis in the Behavioral Sciences'. Harvard University. Retrieved 12 June 2017.
- ^Werbos, Paul (1982). 'Applications of advances in nonlinear sensitivity analysis'(PDF). System modeling and optimization. Springer. pp. 762–770.
- ^ abLeCun et al., 'Backpropagation Applied to Handwritten Zip Code Recognition,' Neural Computation, 1, pp. 541–551, 1989.
- ^J. Weng, N. Ahuja and T. S. Huang, 'Cresceptron: a self-organizing neural network which grows adaptively,' Proc. International Joint Conference on Neural Networks, Baltimore, Maryland, vol I, pp. 576-581, June, 1992.
- ^J. Weng, N. Ahuja and T. S. Huang, 'Learning recognition and segmentation of 3-D objects from 2-D images,' Proc. 4th International Conf. Computer Vision, Berlin, Germany, pp. 121-128, May, 1993.
- ^J. Weng, N. Ahuja and T. S. Huang, 'Learning recognition and segmentation using the Cresceptron,' International Journal of Computer Vision, vol. 25, no. 2, pp. 105-139, Nov. 1997.
- ^de Carvalho, Andre C. L. F.; Fairhurst, Mike C.; Bisset, David (1994-08-08). 'An integrated Boolean neural network for pattern classification'. Pattern Recognition Letters. 15 (8): 807–813. doi:10.1016/0167-8655(94)90009-4.
- ^Hinton, Geoffrey E.; Dayan, Peter; Frey, Brendan J.; Neal, Radford (1995-05-26). 'The wake-sleep algorithm for unsupervised neural networks'. Science. 268 (5214): 1158–1161. Bibcode:1995Sci..268.1158H. doi:10.1126/science.7761831.
- ^ abS. Hochreiter., 'Untersuchungen zu dynamischen neuronalen Netzen,' Diploma thesis. Institut f. Informatik, Technische Univ. Munich. Advisor: J. Schmidhuber, 1991.
- ^Hochreiter, S.; et al. (15 January 2001). 'Gradient flow in recurrent nets: the difficulty of learning long-term dependencies'. In Kolen, John F.; Kremer, Stefan C. (eds.). A Field Guide to Dynamical Recurrent Networks. John Wiley & Sons. ISBN978-0-7803-5369-5.
- ^Morgan, Nelson; Bourlard, Hervé; Renals, Steve; Cohen, Michael; Franco, Horacio (1993-08-01). 'Hybrid neural network/hidden markov model systems for continuous speech recognition'. International Journal of Pattern Recognition and Artificial Intelligence. 07 (4): 899–916. doi:10.1142/s0218001493000455. ISSN0218-0014.
- ^Robinson, T. (1992). 'A real-time recurrent error propagation network word recognition system'. ICASSP: 617–620.
- ^Waibel, A.; Hanazawa, T.; Hinton, G.; Shikano, K.; Lang, K. J. (March 1989). 'Phoneme recognition using time-delay neural networks'. IEEE Transactions on Acoustics, Speech, and Signal Processing. 37 (3): 328–339. doi:10.1109/29.21701. ISSN0096-3518.
- ^Baker, J.; Deng, Li; Glass, Jim; Khudanpur, S.; Lee, C.-H.; Morgan, N.; O'Shaughnessy, D. (2009). 'Research Developments and Directions in Speech Recognition and Understanding, Part 1'. IEEE Signal Processing Magazine. 26 (3): 75–80. Bibcode:2009ISPM..26..75B. doi:10.1109/msp.2009.932166.
- ^Bengio, Y. (1991). 'Artificial Neural Networks and their Application to Speech/Sequence Recognition'. McGill University Ph.D. thesis.
- ^Deng, L.; Hassanein, K.; Elmasry, M. (1994). 'Analysis of correlation structure for a neural predictive model with applications to speech recognition'. Neural Networks. 7 (2): 331–339. doi:10.1016/0893-6080(94)90027-2.
- ^ abHeck, L.; Konig, Y.; Sonmez, M.; Weintraub, M. (2000). 'Robustness to Telephone Handset Distortion in Speaker Recognition by Discriminative Feature Design'. Speech Communication. 31 (2): 181–192. doi:10.1016/s0167-6393(99)00077-1.
- ^'Acoustic Modeling with Deep Neural Networks Using Raw Time Signal for LVCSR (PDF Download Available)'. ResearchGate. Retrieved 2017-06-14.
- ^ abcHochreiter, Sepp; Schmidhuber, Jürgen (1997-11-01). 'Long Short-Term Memory'. Neural Computation. 9 (8): 1735–1780. doi:10.1162/neco.1997.9.8.1735. ISSN0899-7667. PMID9377276.
- ^ abGraves, Alex; Eck, Douglas; Beringer, Nicole; Schmidhuber, Jürgen (2003). 'Biologically Plausible Speech Recognition with LSTM Neural Nets'(PDF). 1st Intl. Workshop on Biologically Inspired Approaches to Advanced Information Technology, Bio-ADIT 2004, Lausanne, Switzerland. pp. 175–184.
- ^ abGraves, Alex; Fernández, Santiago; Gomez, Faustino (2006). 'Connectionist temporal classification: Labelling unsegmented sequence data with recurrent neural networks'. Proceedings of the International Conference on Machine Learning, ICML 2006: 369–376. CiteSeerX10.1.1.75.6306.
- ^Santiago Fernandez, Alex Graves, and Jürgen Schmidhuber (2007). An application of recurrent neural networks to discriminative keyword spotting. Proceedings of ICANN (2), pp. 220–229.
- ^ abSak, Haşim; Senior, Andrew; Rao, Kanishka; Beaufays, Françoise; Schalkwyk, Johan (September 2015). 'Google voice search: faster and more accurate'.
- ^Hinton, Geoffrey E. (2007-10-01). 'Learning multiple layers of representation'. Trends in Cognitive Sciences. 11 (10): 428–434. doi:10.1016/j.tics.2007.09.004. ISSN1364-6613. PMID17921042.
- ^Hinton, G. E.; Osindero, S.; Teh, Y. W. (2006). 'A Fast Learning Algorithm for Deep Belief Nets'(PDF). Neural Computation. 18 (7): 1527–1554. doi:10.1162/neco.2006.18.7.1527. PMID16764513.
- ^Bengio, Yoshua (2012). 'Practical recommendations for gradient-based training of deep architectures'. arXiv:1206.5533 [cs.LG].
- ^G. E. Hinton., 'Learning multiple layers of representation,' Trends in Cognitive Sciences, 11, pp. 428–434, 2007.
- ^ abcHinton, G.; Deng, L.; Yu, D.; Dahl, G.; Mohamed, A.; Jaitly, N.; Senior, A.; Vanhoucke, V.; Nguyen, P.; Sainath, T.; Kingsbury, B. (2012). 'Deep Neural Networks for Acoustic Modeling in Speech Recognition --- The shared views of four research groups'. IEEE Signal Processing Magazine. 29 (6): 82–97. doi:10.1109/msp.2012.2205597.
- ^Deng, Li; Hinton, Geoffrey; Kingsbury, Brian (1 May 2013). 'New types of deep neural network learning for speech recognition and related applications: An overview' – via research.microsoft.com.
- ^Deng, L.; Li, J.; Huang, J. T.; Yao, K.; Yu, D.; Seide, F.; Seltzer, M.; Zweig, G.; He, X. (May 2013). 'Recent advances in deep learning for speech research at Microsoft'. 2013 IEEE International Conference on Acoustics, Speech and Signal Processing: 8604–8608. doi:10.1109/icassp.2013.6639345. ISBN978-1-4799-0356-6.
- ^Sak, Hasim; Senior, Andrew; Beaufays, Francoise (2014). 'Long Short-Term Memory recurrent neural network architectures for large scale acoustic modeling'(PDF).
- ^Li, Xiangang; Wu, Xihong (2014). 'Constructing Long Short-Term Memory based Deep Recurrent Neural Networks for Large Vocabulary Speech Recognition'. arXiv:1410.4281 [cs.CL].
- ^Zen, Heiga; Sak, Hasim (2015). 'Unidirectional Long Short-Term Memory Recurrent Neural Network with Recurrent Output Layer for Low-Latency Speech Synthesis'(PDF). Google.com. ICASSP. pp. 4470–4474.
- ^Deng, L.; Abdel-Hamid, O.; Yu, D. (2013). 'A deep convolutional neural network using heterogeneous pooling for trading acoustic invariance with phonetic confusion'(PDF). Google.com. ICASSP.
- ^ abSainath, T. N.; Mohamed, A. r; Kingsbury, B.; Ramabhadran, B. (May 2013). 'Deep convolutional neural networks for LVCSR'. 2013 IEEE International Conference on Acoustics, Speech and Signal Processing: 8614–8618. doi:10.1109/icassp.2013.6639347. ISBN978-1-4799-0356-6.
- ^Yann LeCun (2016). Slides on Deep Learning Online
- ^ abcNIPS Workshop: Deep Learning for Speech Recognition and Related Applications, Whistler, BC, Canada, Dec. 2009 (Organizers: Li Deng, Geoff Hinton, D. Yu).
- ^ abKeynote talk: Recent Developments in Deep Neural Networks. ICASSP, 2013 (by Geoff Hinton).
- ^D. Yu, L. Deng, G. Li, and F. Seide (2011). 'Discriminative pretraining of deep neural networks,' U.S. Patent Filing.
- ^ abcDeng, L.; Hinton, G.; Kingsbury, B. (2013). 'New types of deep neural network learning for speech recognition and related applications: An overview (ICASSP)'(PDF).
- ^ abcYu, D.; Deng, L. (2014). Automatic Speech Recognition: A Deep Learning Approach (Publisher: Springer). ISBN978-1-4471-5779-3.
- ^'Deng receives prestigious IEEE Technical Achievement Award - Microsoft Research'. Microsoft Research. 3 December 2015.
- ^ abLi, Deng (September 2014). 'Keynote talk: 'Achievements and Challenges of Deep Learning - From Speech Analysis and Recognition To Language and Multimodal Processing''. Interspeech.
- ^Yu, D.; Deng, L. (2010). 'Roles of Pre-Training and Fine-Tuning in Context-Dependent DBN-HMMs for Real-World Speech Recognition'. NIPS Workshop on Deep Learning and Unsupervised Feature Learning.
- ^Seide, F.; Li, G.; Yu, D. (2011). 'Conversational speech transcription using context-dependent deep neural networks'. Interspeech.
- ^Deng, Li; Li, Jinyu; Huang, Jui-Ting; Yao, Kaisheng; Yu, Dong; Seide, Frank; Seltzer, Mike; Zweig, Geoff; He, Xiaodong (2013-05-01). 'Recent Advances in Deep Learning for Speech Research at Microsoft'. Microsoft Research.
- ^'Nvidia CEO bets big on deep learning and VR'. Venture Beat. April 5, 2016.
- ^'From not working to neural networking'. The Economist.
- ^ abOh, K.-S.; Jung, K. (2004). 'GPU implementation of neural networks'. Pattern Recognition. 37 (6): 1311–1314. doi:10.1016/j.patcog.2004.01.013.
- ^ abChellapilla, K., Puri, S., and Simard, P. (2006). High performance convolutional neural networks for document processing. International Workshop on Frontiers in Handwriting Recognition.
- ^Cireşan, Dan Claudiu; Meier, Ueli; Gambardella, Luca Maria; Schmidhuber, Jürgen (2010-09-21). 'Deep, Big, Simple Neural Nets for Handwritten Digit Recognition'. Neural Computation. 22 (12): 3207–3220. arXiv:1003.0358. doi:10.1162/neco_a_00052. ISSN0899-7667. PMID20858131.
- ^Raina, Rajat; Madhavan, Anand; Ng, Andrew Y. (2009). 'Large-scale Deep Unsupervised Learning Using Graphics Processors'. Proceedings of the 26th Annual International Conference on Machine Learning. ICML '09. New York, NY, USA: ACM: 873–880. CiteSeerX10.1.1.154.372. doi:10.1145/1553374.1553486. ISBN9781605585161.
- ^Sze, Vivienne; Chen, Yu-Hsin; Yang, Tien-Ju; Emer, Joel (2017). 'Efficient Processing of Deep Neural Networks: A Tutorial and Survey'. arXiv:1703.09039 [cs.CV].
- ^ ab'Announcement of the winners of the Merck Molecular Activity Challenge'.
- ^ ab'Multi-task Neural Networks for QSAR Predictions | Data Science Association'. www.datascienceassn.org. Retrieved 2017-06-14.
- ^ ab'Toxicology in the 21st century Data Challenge'
- ^ ab'NCATS Announces Tox21 Data Challenge Winners'.
- ^ ab'Archived copy'. Archived from the original on 2015-02-28. Retrieved 2015-03-05.CS1 maint: Archived copy as title (link)
- ^Ciresan, D. C.; Meier, U.; Masci, J.; Gambardella, L. M.; Schmidhuber, J. (2011). 'Flexible, High Performance Convolutional Neural Networks for Image Classification'(PDF). International Joint Conference on Artificial Intelligence. doi:10.5591/978-1-57735-516-8/ijcai11-210.
- ^Ciresan, Dan; Giusti, Alessandro; Gambardella, Luca M.; Schmidhuber, Juergen (2012). Pereira, F.; Burges, C. J. C.; Bottou, L.; Weinberger, K. Q. (eds.). Advances in Neural Information Processing Systems 25(PDF). Curran Associates, Inc. pp. 2843–2851.
- ^Ciresan, D.; Giusti, A.; Gambardella, L.M.; Schmidhuber, J. (2013). 'Mitosis Detection in Breast Cancer Histology Images using Deep Neural Networks'. Proceedings MICCAI. Lecture Notes in Computer Science. 7908: 411–418. doi:10.1007/978-3-642-40763-5_51. ISBN978-3-642-38708-1.
- ^'The Wolfram Language Image Identification Project'. www.imageidentify.com. Retrieved 2017-03-22.
- ^Vinyals, Oriol; Toshev, Alexander; Bengio, Samy; Erhan, Dumitru (2014). 'Show and Tell: A Neural Image Caption Generator'. arXiv:1411.4555 [cs.CV]..
- ^Fang, Hao; Gupta, Saurabh; Iandola, Forrest; Srivastava, Rupesh; Deng, Li; Dollár, Piotr; Gao, Jianfeng; He, Xiaodong; Mitchell, Margaret; Platt, John C; Lawrence Zitnick, C; Zweig, Geoffrey (2014). 'From Captions to Visual Concepts and Back'. arXiv:1411.4952 [cs.CV]..
- ^Kiros, Ryan; Salakhutdinov, Ruslan; Zemel, Richard S (2014). 'Unifying Visual-Semantic Embeddings with Multimodal Neural Language Models'. arXiv:1411.2539 [cs.LG]..
- ^Zhong, Sheng-hua; Liu, Yan; Liu, Yang (2011). 'Bilinear Deep Learning for Image Classification'. Proceedings of the 19th ACM International Conference on Multimedia. MM '11. New York, NY, USA: ACM: 343–352. doi:10.1145/2072298.2072344. ISBN9781450306164.
- ^'Why Deep Learning Is Suddenly Changing Your Life'. Fortune. 2016. Retrieved 13 April 2018.
- ^Silver, David; Huang, Aja; Maddison, Chris J.; Guez, Arthur; Sifre, Laurent; Driessche, George van den; Schrittwieser, Julian; Antonoglou, Ioannis; Panneershelvam, Veda (January 2016). 'Mastering the game of Go with deep neural networks and tree search'. Nature. 529 (7587): 484–489. Bibcode:2016Natur.529.484S. doi:10.1038/nature16961. ISSN1476-4687. PMID26819042.
- ^Szegedy, Christian; Toshev, Alexander; Erhan, Dumitru (2013). 'Deep neural networks for object detection'. Advances in Neural Information Processing Systems.
- ^Hof, Robert D. 'Is Artificial Intelligence Finally Coming into Its Own?'. MIT Technology Review. Retrieved 2018-07-10.
- ^ abGers, Felix A.; Schmidhuber, Jürgen (2001). 'LSTM Recurrent Networks Learn Simple Context Free and Context Sensitive Languages'. IEEE Trans. Neural Netw. 12 (6): 1333–1340. doi:10.1109/72.963769. PMID18249962.
- ^ abcSutskever, L.; Vinyals, O.; Le, Q. (2014). 'Sequence to Sequence Learning with Neural Networks'(PDF). Proc. NIPS.
- ^ abJozefowicz, Rafal; Vinyals, Oriol; Schuster, Mike; Shazeer, Noam; Wu, Yonghui (2016). 'Exploring the Limits of Language Modeling'. arXiv:1602.02410 [cs.CL].
- ^ abGillick, Dan; Brunk, Cliff; Vinyals, Oriol; Subramanya, Amarnag (2015). 'Multilingual Language Processing from Bytes'. arXiv:1512.00103 [cs.CL].
- ^Mikolov, T.; et al. (2010). 'Recurrent neural network based language model'(PDF). Interspeech.
- ^ ab'Learning Precise Timing with LSTM Recurrent Networks (PDF Download Available)'. ResearchGate. Retrieved 2017-06-13.
- ^LeCun, Y.; et al. (1998). 'Gradient-based learning applied to document recognition'. Proceedings of the IEEE. 86 (11): 2278–2324. doi:10.1109/5.726791.
- ^Bengio, Y.; Boulanger-Lewandowski, N.; Pascanu, R. (May 2013). 'Advances in optimizing recurrent networks'. 2013 IEEE International Conference on Acoustics, Speech and Signal Processing: 8624–8628. arXiv:1212.0901. CiteSeerX10.1.1.752.9151. doi:10.1109/icassp.2013.6639349. ISBN978-1-4799-0356-6.
- ^Dahl, G.; et al. (2013). 'Improving DNNs for LVCSR using rectified linear units and dropout'(PDF). ICASSP.
- ^'Data Augmentation - deeplearning.ai | Coursera'. Coursera. Retrieved 2017-11-30.
- ^Hinton, G. E. (2010). 'A Practical Guide to Training Restricted Boltzmann Machines'. Tech. Rep. UTML TR 2010-003.
- ^You, Yang; Buluç, Aydın; Demmel, James (November 2017). 'Scaling deep learning on GPU and knights landing clusters'. SC '17, ACM. Retrieved 5 March 2018.
- ^Viebke, André; Memeti, Suejb; Pllana, Sabri; Abraham, Ajith (March 2017). 'CHAOS: a parallelization scheme for training convolutional neural networks on Intel Xeon Phi'. The Journal of Supercomputing. 75: 197–227. doi:10.1007/s11227-017-1994-x.
- ^Ting Qin, et al. 'A learning algorithm of CMAC based on RLS.' Neural Processing Letters 19.1 (2004): 49-61.
- ^Ting Qin, et al. 'Continuous CMAC-QRLS and its systolic array.' Neural Processing Letters 22.1 (2005): 1-16.
- ^TIMIT Acoustic-Phonetic Continuous Speech Corpus Linguistic Data Consortium, Philadelphia.
- ^Robinson, Tony (30 September 1991). 'Several Improvements to a Recurrent Error Propagation Network Phone Recognition System'. Cambridge University Engineering Department Technical Report. CUED/F-INFENG/TR82. doi:10.13140/RG.2.2.15418.90567.
- ^Abdel-Hamid, O.; et al. (2014). 'Convolutional Neural Networks for Speech Recognition'. IEEE/ACM Transactions on Audio, Speech, and Language Processing. 22 (10): 1533–1545. doi:10.1109/taslp.2014.2339736.
- ^Deng, L.; Platt, J. (2014). 'Ensemble Deep Learning for Speech Recognition'(PDF). Proc. Interspeech.
- ^Tóth, Laszló (2015). 'Phone Recognition with Hierarchical Convolutional Deep Maxout Networks'(PDF). EURASIP Journal on Audio, Speech, and Music Processing. 2015. doi:10.1186/s13636-015-0068-3.
- ^'How Skype Used AI to Build Its Amazing New Language Translator | WIRED'. www.wired.com. Retrieved 2017-06-14.
- ^Hannun, Awni; Case, Carl; Casper, Jared; Catanzaro, Bryan; Diamos, Greg; Elsen, Erich; Prenger, Ryan; Satheesh, Sanjeev; Sengupta, Shubho; Coates, Adam; Ng, Andrew Y (2014). 'Deep Speech: Scaling up end-to-end speech recognition'. arXiv:1412.5567 [cs.CL].
- ^'Plenary presentation at ICASSP-2016'(PDF).
- ^'MNIST handwritten digit database, Yann LeCun, Corinna Cortes and Chris Burges'. yann.lecun.com.
- ^Cireşan, Dan; Meier, Ueli; Masci, Jonathan; Schmidhuber, Jürgen (August 2012). 'Multi-column deep neural network for traffic sign classification'. Neural Networks. Selected Papers from IJCNN 2011. 32: 333–338. CiteSeerX10.1.1.226.8219. doi:10.1016/j.neunet.2012.02.023. PMID22386783.
- ^Nvidia Demos a Car Computer Trained with 'Deep Learning' (2015-01-06), David Talbot, MIT Technology Review
- ^G. W. Smith; Frederic Fol Leymarie (10 April 2017). 'The Machine as Artist: An Introduction'. Arts. Retrieved 4 October 2017.
- ^Blaise Agüera y Arcas (29 September 2017). 'Art in the Age of Machine Intelligence'. Arts. Retrieved 4 October 2017.
- ^Bengio, Yoshua; Ducharme, Réjean; Vincent, Pascal; Janvin, Christian (March 2003). 'A Neural Probabilistic Language Model'. J. Mach. Learn. Res. 3: 1137–1155. ISSN1532-4435.
- ^Goldberg, Yoav; Levy, Omar (2014). 'word2vec Explained: Deriving Mikolov et al.'s Negative-Sampling Word-Embedding Method'. arXiv:1402.3722 [cs.CL].
- ^ abSocher, Richard; Manning, Christopher. 'Deep Learning for NLP'(PDF). Retrieved 26 October 2014.
- ^Socher, Richard; Bauer, John; Manning, Christopher; Ng, Andrew (2013). 'Parsing With Compositional Vector Grammars'(PDF). Proceedings of the ACL 2013 Conference.
- ^Socher, Richard (2013). 'Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank'(PDF).
- ^Shen, Yelong; He, Xiaodong; Gao, Jianfeng; Deng, Li; Mesnil, Gregoire (2014-11-01). 'A Latent Semantic Model with Convolutional-Pooling Structure for Information Retrieval'. Microsoft Research.
- ^Huang, Po-Sen; He, Xiaodong; Gao, Jianfeng; Deng, Li; Acero, Alex; Heck, Larry (2013-10-01). 'Learning Deep Structured Semantic Models for Web Search using Clickthrough Data'. Microsoft Research.
- ^Mesnil, G.; Dauphin, Y.; Yao, K.; Bengio, Y.; Deng, L.; Hakkani-Tur, D.; He, X.; Heck, L.; Tur, G.; Yu, D.; Zweig, G. (2015). 'Using recurrent neural networks for slot filling in spoken language understanding'. IEEE Transactions on Audio, Speech, and Language Processing. 23 (3): 530–539. doi:10.1109/taslp.2014.2383614.
- ^ abGao, Jianfeng; He, Xiaodong; Yih, Scott Wen-tau; Deng, Li (2014-06-01). 'Learning Continuous Phrase Representations for Translation Modeling'. Microsoft Research.
- ^Brocardo, Marcelo Luiz; Traore, Issa; Woungang, Isaac; Obaidat, Mohammad S. (2017). 'Authorship verification using deep belief network systems'. International Journal of Communication Systems. 30 (12): e3259. doi:10.1002/dac.3259.
- ^'Deep Learning for Natural Language Processing: Theory and Practice (CIKM2014 Tutorial) - Microsoft Research'. Microsoft Research. Retrieved 2017-06-14.
- ^Turovsky, Barak (November 15, 2016). 'Found in translation: More accurate, fluent sentences in Google Translate'. The Keyword Google Blog. Retrieved March 23, 2017.
- ^ abcdSchuster, Mike; Johnson, Melvin; Thorat, Nikhil (November 22, 2016). 'Zero-Shot Translation with Google's Multilingual Neural Machine Translation System'. Google Research Blog. Retrieved March 23, 2017.
- ^Sepp Hochreiter; Jürgen Schmidhuber (1997). 'Long short-term memory'. Neural Computation. 9 (8): 1735–1780. doi:10.1162/neco.1997.9.8.1735. PMID9377276.
- ^Felix A. Gers; Jürgen Schmidhuber; Fred Cummins (2000). 'Learning to Forget: Continual Prediction with LSTM'. Neural Computation. 12 (10): 2451–2471. CiteSeerX10.1.1.55.5709. doi:10.1162/089976600300015015.
- ^Wu, Yonghui; Schuster, Mike; Chen, Zhifeng; Le, Quoc V; Norouzi, Mohammad; Macherey, Wolfgang; Krikun, Maxim; Cao, Yuan; Gao, Qin; Macherey, Klaus; Klingner, Jeff; Shah, Apurva; Johnson, Melvin; Liu, Xiaobing; Kaiser, Łukasz; Gouws, Stephan; Kato, Yoshikiyo; Kudo, Taku; Kazawa, Hideto; Stevens, Keith; Kurian, George; Patil, Nishant; Wang, Wei; Young, Cliff; Smith, Jason; Riesa, Jason; Rudnick, Alex; Vinyals, Oriol; Corrado, Greg; et al. (2016). 'Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation'. arXiv:1609.08144 [cs.CL].
- ^'An Infusion of AI Makes Google Translate More Powerful Than Ever.' Cade Metz, WIRED, Date of Publication: 09.27.16. https://www.wired.com/2016/09/google-claims-ai-breakthrough-machine-translation/
- ^ abBoitet, Christian; Blanchon, Hervé; Seligman, Mark; Bellynck, Valérie (2010). 'MT on and for the Web'(PDF). Retrieved December 1, 2016.
- ^Arrowsmith, J; Miller, P (2013). 'Trial watch: Phase II and phase III attrition rates 2011-2012'. Nature Reviews Drug Discovery. 12 (8): 569. doi:10.1038/nrd4090. PMID23903212.
- ^Verbist, B; Klambauer, G; Vervoort, L; Talloen, W; The Qstar, Consortium; Shkedy, Z; Thas, O; Bender, A; Göhlmann, H. W.; Hochreiter, S (2015). 'Using transcriptomics to guide lead optimization in drug discovery projects: Lessons learned from the QSTAR project'. Drug Discovery Today. 20 (5): 505–513. doi:10.1016/j.drudis.2014.12.014. PMID25582842.
- ^Wallach, Izhar; Dzamba, Michael; Heifets, Abraham (2015-10-09). 'AtomNet: A Deep Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug Discovery'. arXiv:1510.02855 [cs.LG].
- ^'Toronto startup has a faster way to discover effective medicines'. The Globe and Mail. Retrieved 2015-11-09.
- ^'Startup Harnesses Supercomputers to Seek Cures'. KQED Future of You. Retrieved 2015-11-09.
- ^'Toronto startup has a faster way to discover effective medicines'.
- ^Tkachenko, Yegor (April 8, 2015). 'Autonomous CRM Control via CLV Approximation with Deep Reinforcement Learning in Discrete and Continuous Action Space'. arXiv:1504.01840 [cs.LG].
- ^van den Oord, Aaron; Dieleman, Sander; Schrauwen, Benjamin (2013). Burges, C. J. C.; Bottou, L.; Welling, M.; Ghahramani, Z.; Weinberger, K. Q. (eds.). Advances in Neural Information Processing Systems 26(PDF). Curran Associates, Inc. pp. 2643–2651.
- ^Elkahky, Ali Mamdouh; Song, Yang; He, Xiaodong (2015-05-01). 'A Multi-View Deep Learning Approach for Cross Domain User Modeling in Recommendation Systems'. Microsoft Research.
- ^Chicco, Davide; Sadowski, Peter; Baldi, Pierre (1 January 2014). Deep Autoencoder Neural Networks for Gene Ontology Annotation Predictions. Proceedings of the 5th ACM Conference on Bioinformatics, Computational Biology, and Health Informatics - BCB '14. ACM. pp. 533–540. doi:10.1145/2649387.2649442. hdl:11311/964622. ISBN9781450328944.
- ^Sathyanarayana, Aarti (2016-01-01). 'Sleep Quality Prediction From Wearable Data Using Deep Learning'. JMIR mHealth and uHealth. 4 (4): e125. doi:10.2196/mhealth.6562. PMC5116102. PMID27815231.
- ^Choi, Edward; Schuetz, Andy; Stewart, Walter F.; Sun, Jimeng (2016-08-13). 'Using recurrent neural network models for early detection of heart failure onset'. Journal of the American Medical Informatics Association. 24 (2): 361–370. doi:10.1093/jamia/ocw112. ISSN1067-5027. PMC5391725. PMID27521897.
- ^'Deep Learning in Healthcare: Challenges and Opportunities'. Medium. 2016-08-12. Retrieved 2018-04-10.
- ^Litjens, Geert; Kooi, Thijs; Bejnordi, Babak Ehteshami; Setio, Arnaud Arindra Adiyoso; Ciompi, Francesco; Ghafoorian, Mohsen; van der Laak, Jeroen A.W.M.; van Ginneken, Bram; Sánchez, Clara I. (December 2017). 'A survey on deep learning in medical image analysis'. Medical Image Analysis. 42: 60–88. doi:10.1016/j.media.2017.07.005.
- ^Forslid, Gustav; Wieslander, Hakan; Bengtsson, Ewert; Wahlby, Carolina; Hirsch, Jan-Michael; Stark, Christina Runow; Sadanandan, Sajith Kecheril (October 2017). 'Deep Convolutional Neural Networks for Detecting Cellular Changes Due to Malignancy'. 2017 IEEE International Conference on Computer Vision Workshops (ICCVW). Venice: IEEE: 82–89. doi:10.1109/ICCVW.2017.18. ISBN9781538610343.
- ^De, Shaunak; Maity, Abhishek; Goel, Vritti; Shitole, Sanjay; Bhattacharya, Avik (2017). 'Predicting the popularity of instagram posts for a lifestyle magazine using deep learning'. 2nd IEEE Conference on Communication Systems, Computing and IT Applications: 174–177. doi:10.1109/CSCITA.2017.8066548. ISBN978-1-5090-4381-1.
- ^Schmidt, Uwe; Roth, Stefan. Shrinkage Fields for Effective Image Restoration(PDF). Computer Vision and Pattern Recognition (CVPR), 2014 IEEE Conference on.
- ^Czech, Tomasz. 'Deep learning: the next frontier for money laundering detection'. Global Banking and Finance Review.
- ^ abc'Army researchers develop new algorithms to train robots'. EurekAlert!. Retrieved 2018-08-29.
- ^Utgoff, P. E.; Stracuzzi, D. J. (2002). 'Many-layered learning'. Neural Computation. 14 (10): 2497–2529. doi:10.1162/08997660260293319. PMID12396572.
- ^Elman, Jeffrey L. (1998). Rethinking Innateness: A Connectionist Perspective on Development. MIT Press. ISBN978-0-262-55030-7.
- ^Shrager, J.; Johnson, MH (1996). 'Dynamic plasticity influences the emergence of function in a simple cortical array'. Neural Networks. 9 (7): 1119–1129. doi:10.1016/0893-6080(96)00033-0. PMID12662587.
- ^Quartz, SR; Sejnowski, TJ (1997). 'The neural basis of cognitive development: A constructivist manifesto'. Behavioral and Brain Sciences. 20 (4): 537–556. CiteSeerX10.1.1.41.7854. doi:10.1017/s0140525x97001581.
- ^S. Blakeslee., 'In brain's early growth, timetable may be critical,' The New York Times, Science Section, pp. B5–B6, 1995.
- ^Mazzoni, P.; Andersen, R. A.; Jordan, M. I. (1991-05-15). 'A more biologically plausible learning rule for neural networks'. Proceedings of the National Academy of Sciences. 88 (10): 4433–4437. Bibcode:1991PNAS..88.4433M. doi:10.1073/pnas.88.10.4433. ISSN0027-8424. PMC51674. PMID1903542.
- ^O'Reilly, Randall C. (1996-07-01). 'Biologically Plausible Error-Driven Learning Using Local Activation Differences: The Generalized Recirculation Algorithm'. Neural Computation. 8 (5): 895–938. doi:10.1162/neco.1996.8.5.895. ISSN0899-7667.
- ^Testolin, Alberto; Zorzi, Marco (2016). 'Probabilistic Models and Generative Neural Networks: Towards an Unified Framework for Modeling Normal and Impaired Neurocognitive Functions'. Frontiers in Computational Neuroscience. 10: 73. doi:10.3389/fncom.2016.00073. ISSN1662-5188. PMC4943066. PMID27468262.
- ^Testolin, Alberto; Stoianov, Ivilin; Zorzi, Marco (September 2017). 'Letter perception emerges from unsupervised deep learning and recycling of natural image features'. Nature Human Behaviour. 1 (9): 657–664. doi:10.1038/s41562-017-0186-2. ISSN2397-3374.
- ^Buesing, Lars; Bill, Johannes; Nessler, Bernhard; Maass, Wolfgang (2011-11-03). 'Neural Dynamics as Sampling: A Model for Stochastic Computation in Recurrent Networks of Spiking Neurons'. PLOS Computational Biology. 7 (11): e1002211. Bibcode:2011PLSCB..7E2211B. doi:10.1371/journal.pcbi.1002211. ISSN1553-7358. PMC3207943. PMID22096452.
- ^Morel, Danielle; Singh, Chandan; Levy, William B. (2018-01-25). 'Linearization of excitatory synaptic integration at no extra cost'. Journal of Computational Neuroscience. 44 (2): 173–188. doi:10.1007/s10827-017-0673-5. ISSN0929-5313. PMID29372434.
- ^Cash, S.; Yuste, R. (February 1999). 'Linear summation of excitatory inputs by CA1 pyramidal neurons'. Neuron. 22 (2): 383–394. doi:10.1016/s0896-6273(00)81098-3. ISSN0896-6273. PMID10069343.
- ^Olshausen, B; Field, D (2004-08-01). 'Sparse coding of sensory inputs'. Current Opinion in Neurobiology. 14 (4): 481–487. doi:10.1016/j.conb.2004.07.007. ISSN0959-4388.
- ^Yamins, Daniel L K; DiCarlo, James J (March 2016). 'Using goal-driven deep learning models to understand sensory cortex'. Nature Neuroscience. 19 (3): 356–365. doi:10.1038/nn.4244. ISSN1546-1726.
- ^Zorzi, Marco; Testolin, Alberto (2018-02-19). 'An emergentist perspective on the origin of number sense'. Phil. Trans. R. Soc. B. 373 (1740): 20170043. doi:10.1098/rstb.2017.0043. ISSN0962-8436. PMC5784047. PMID29292348.
- ^Güçlü, Umut; van Gerven, Marcel A. J. (2015-07-08). 'Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Ventral Stream'. Journal of Neuroscience. 35 (27): 10005–10014. arXiv:1411.6422. doi:10.1523/jneurosci.5023-14.2015. PMID26157000.
- ^Metz, C. (12 December 2013). 'Facebook's 'Deep Learning' Guru Reveals the Future of AI'. Wired.
- ^'Google AI algorithm masters ancient game of Go'. Nature News & Comment. Retrieved 2016-01-30.
- ^Silver, David; Huang, Aja; Maddison, Chris J.; Guez, Arthur; Sifre, Laurent; Driessche, George van den; Schrittwieser, Julian; Antonoglou, Ioannis; Panneershelvam, Veda; Lanctot, Marc; Dieleman, Sander; Grewe, Dominik; Nham, John; Kalchbrenner, Nal; Sutskever, Ilya; Lillicrap, Timothy; Leach, Madeleine; Kavukcuoglu, Koray; Graepel, Thore; Hassabis, Demis (28 January 2016). 'Mastering the game of Go with deep neural networks and tree search'. Nature. 529 (7587): 484–489. Bibcode:2016Natur.529.484S. doi:10.1038/nature16961. ISSN0028-0836. PMID26819042.
- ^'A Google DeepMind Algorithm Uses Deep Learning and More to Master the Game of Go | MIT Technology Review'. MIT Technology Review. Retrieved 2016-01-30.
- ^'Blippar Demonstrates New Real-Time Augmented Reality App'. TechCrunch.
- ^'TAMER: Training an Agent Manually via Evaluative Reinforcement - IEEE Conference Publication'. ieeexplore.ieee.org. Retrieved 2018-08-29.
- ^'Talk to the Algorithms: AI Becomes a Faster Learner'. governmentciomedia.com. Retrieved 2018-08-29.
- ^Marcus, Gary (2018-01-14). 'In defense of skepticism about deep learning'. Gary Marcus. Retrieved 2018-10-11.
- ^Knight, Will (2017-03-14). 'DARPA is funding projects that will try to open up AI's black boxes'. MIT Technology Review. Retrieved 2017-11-02.
- ^Marcus, Gary (November 25, 2012). 'Is 'Deep Learning' a Revolution in Artificial Intelligence?'. The New Yorker. Retrieved 2017-06-14.
- ^Smith, G. W. (March 27, 2015). 'Art and Artificial Intelligence'. ArtEnt. Archived from the original on June 25, 2017. Retrieved March 27, 2015.CS1 maint: BOT: original-url status unknown (link)
- ^Mellars, Paul (February 1, 2005). 'The Impossible Coincidence: A Single-Species Model for the Origins of Modern Human Behavior in Europe'(PDF). Evolutionary Anthropology: Issues, News, and Reviews. Retrieved April 5, 2017.
- ^Alexander Mordvintsev; Christopher Olah; Mike Tyka (June 17, 2015). 'Inceptionism: Going Deeper into Neural Networks'. Google Research Blog. Retrieved June 20, 2015.
- ^Alex Hern (June 18, 2015). 'Yes, androids do dream of electric sheep'. The Guardian. Retrieved June 20, 2015.
- ^ abcGoertzel, Ben (2015). 'Are there Deep Reasons Underlying the Pathologies of Today's Deep Learning Algorithms?'(PDF).
- ^Nguyen, Anh; Yosinski, Jason; Clune, Jeff (2014). 'Deep Neural Networks are Easily Fooled: High Confidence Predictions for Unrecognizable Images'. arXiv:1412.1897 [cs.CV].
- ^Szegedy, Christian; Zaremba, Wojciech; Sutskever, Ilya; Bruna, Joan; Erhan, Dumitru; Goodfellow, Ian; Fergus, Rob (2013). 'Intriguing properties of neural networks'. arXiv:1312.6199 [cs.CV].
- ^Zhu, S.C.; Mumford, D. (2006). 'A stochastic grammar of images'. Found. Trends Comput. Graph. Vis. 2 (4): 259–362. CiteSeerX10.1.1.681.2190. doi:10.1561/0600000018.
- ^Miller, G. A., and N. Chomsky. 'Pattern conception.' Paper for Conference on pattern detection, University of Michigan. 1957.
- ^Eisner, Jason. 'Deep Learning of Recursive Structure: Grammar Induction'.
- ^ abcde'AI Is Easy to Fool—Why That Needs to Change'. Singularity Hub. 2017-10-10. Retrieved 2017-10-11.
- ^Gibney, Elizabeth (2017). 'The scientist who spots fake videos'. Nature. doi:10.1038/nature.2017.22784.
Further reading[edit]
- Goodfellow, Ian; Bengio, Yoshua; Courville, Aaron (2016). Deep Learning. MIT Press. ISBN978-0-26203561-3, introductory textbook.
Retrieved from 'https://en.wikipedia.org/w/index.php?title=Deep_learning&oldid=899278872'
Shimer College students learning to cook by cooking, 1942.
Experiential learning is the process of learning through experience, and is more specifically defined as 'learning through reflection on doing'.[1]Hands-on learning is a form of experiential learning but does not necessarily involve students reflecting on their product.[2][3][4] Experiential learning is distinct from rote or didactic learning, in which the learner plays a comparatively passive role.[5] It is related to, but not synonymous with, other forms of active learning such as action learning, adventure learning, free-choice learning, cooperative learning, service-learning, and situated learning.[6]
Experiential learning is often used synonymously with the term 'experiential education', but while experiential education is a broader philosophy of education, experiential learning considers the individual learning process.[7] As such, compared to experiential education, experiential learning is concerned with more concrete issues related to the learner and the learning context.
The general concept of learning through experience is ancient. Around 350 BCE, Aristotle wrote in the Nicomachean Ethics 'for the things we have to learn before we can do them, we learn by doing them'.[8] But as an articulated educational approach, experiential learning is of much more recent vintage. Beginning in the 1970s, David A. Kolb helped to develop the modern theory of experiential learning, drawing heavily on the work of John Dewey, Kurt Lewin, and Jean Piaget.[9]
Experiential learning has significant teaching advantages. Peter Senge, author of The Fifth Discipline (1990), states that teaching is of utmost importance to motivate people. Learning only has good effects when learners have the desire to absorb the knowledge. Therefore, experiential learning requires the showing of directions for learners.[10]
Experiential learning entails a hands-on approach to learning that moves away from just the teacher at the front of the room emparting and transferring their knowledge to students. It makes learning an experience that moves beyond the classroom and strives to bring a more involved way of learning.
- 1Kolb's experiential learning model
- 8See also
Kolb's experiential learning model[edit]
Experiential learning focuses on the learning process for the individual. One example of experiential learning is going to the zoo and learning through observation and interaction with the zoo environment, as opposed to reading about animals from a book. Thus, one makes discoveries and experiments with knowledge firsthand, instead of hearing or reading about others' experiences. Likewise, in business school, internship, and job-shadowing, opportunities in a student’s field of interest can provide valuable experiential learning which contributes significantly to the student’s overall understanding of the real-world environment.[11]
Define Aprendizaje Significativo
A third example of experiential learning involves learning how to ride a bike,[12] a process which can illustrate the four-step experiential learning model (ELM) as set forth by Kolb[13] and outlined in Figure 1 below. Following this example, in the 'concrete experience' stage, the learner physically experiences the bike in the 'here-and-now'.[14] This experience forms 'the basis for observation and reflection' and the learner has the opportunity to consider what is working or failing (reflective observation), and to think about ways to improve on the next attempt made at riding (abstract conceptualization). Every new attempt to ride is informed by a cyclical pattern of previous experience, thought and reflection (active experimentation).[14]
Figure 1 – David Kolb's Experiential Learning Model (ELM)[15]
→ | Concrete Experience | ↓ |
Active Experimentation | Reflective Observation | |
↑ | Abstract Conceptualization | ← |
Elements[edit]
Experiential learning can exist without a teacher and relates solely to the meaning-making process of the individual's direct experience. However, though the gaining of knowledge is an inherent process that occurs naturally, a genuine learning experience requires certain elements.[6] According to Kolb, knowledge is continuously gained through both personal and environmental experiences.[16] Kolb states that in order to gain genuine knowledge from an experience, the learner must have four abilities:
- The learner must be willing to be actively involved in the experience;
- The learner must be able to reflect on the experience;
- The learner must possess and use analytical skills to conceptualize the experience; and
- The learner must possess decision making and problem solving skills in order to use the new ideas gained from the experience.
Implementation[edit]
Experiential learning requires self-initiative, an 'intention to learn' and an 'active phase of learning'.[17] Kolb's cycle of experiential learning can be used as a framework for considering the different stages involved.[18] Jennifer A. Moon has elaborated on this cycle to argue that experiential learning is most effective when it involves: 1) a 'reflective learning phase' 2) a phase of learning resulting from the actions inherent to experiential learning, and 3) 'a further phase of learning from feedback'.[17] This process of learning can result in 'changes in judgment, feeling or skills' for the individual[19] and can provide direction for the 'making of judgments as a guide to choice and action'.[20]
Most educators understand the important role experience plays in the learning process. The role of emotion and feelings in learning from experience has been recognised as an important part of experiential learning.[17] While those factors may improve the likelihood of experiential learning occurring, it can occur without them. Rather, what is vital in experiential learning is that the individual is encouraged to directly involve themselves in the experience, and then to reflect on their experiences using analytic skills, in order that they gain a better understanding of the new knowledge and retain the information for a longer time.
Reflection is a crucial part of the experiential learning process, and like experiential learning itself, it can be facilitated or independent. Dewey wrote that 'successive portions of reflective thought grow out of one another and support one another', creating a scaffold for further learning, and allowing for further experiences and reflection.[21] This reinforces the fact that experiential learning and reflective learning are iterative processes, and the learning builds and develops with further reflection and experience. Facilitation of experiential learning and reflection is challenging, but 'a skilled facilitator, asking the right questions and guiding reflective conversation before, during, and after an experience, can help open a gateway to powerful new thinking and learning'.[22] Jacobson and Ruddy, building on Kolb's four-stage Experiential Learning Model[14] and Pfeiffer and Jones's five stage Experiential Learning Cycle,[23] took these theoretical frameworks and created a simple, practical questioning model for facilitators to use in promoting critical reflection in experiential learning. Their '5 Questions' model is as follows:[22]
- Did you notice?
- Why did that happen?
- Does that happen in life?
- Why does that happen?
- How can you use that?
These questions are posed by the facilitator after an experience, and gradually lead the group towards a critical reflection on their experience, and an understanding of how they can apply the learning to their own life.[22] Although the questions are simple, they allow a relatively inexperienced facilitator to apply the theories of Kolb, Pfeiffer, and Jones, and deepen the learning of the group.
Define Aprendizaje Autonomo
While it is the learner's experience that is most important to the learning process, it is also important not to forget the wealth of experience a good facilitator also brings to the situation. However, while a facilitator, or 'teacher', may improve the likelihood of experiential learning occurring, a facilitator is not essential to experiential learning. Rather, the mechanism of experiential learning is the learner's reflection on experiences using analytic skills. This can occur without the presence of a facilitator, meaning that experiential learning is not defined by the presence of a facilitator. Yet, by considering experiential learning in developing course or program content, it provides an opportunity to develop a framework for adapting varying teaching/learning techniques into the classroom.[24]
In schools[edit]
Experiential learning is supported in different school organizational models and learning environments.
- Think Global School is a four-year traveling high school that holds classes in a new country each term. Students engage in experiential learning through activities such as workshops, cultural exchanges, museum tours, and nature expeditions.
- The Dawson School in Boulder, Colorado, devotes two weeks of each school year to experiential learning, with students visiting surrounding states to engage in community service, visit museums and scientific institutions, and engage in activities such as mountain biking, backpacking, and canoeing.
- In the ELENA-Project, the follow-up project of 'animals live', experiential learning with living animals will be developed. Together with project partners from Romania, Hungary and Georgia, the Bavarian Academy of Nature Conservation and Landscape Management in Germany brings living animals in the lessons of European schools. The aim is to brief children for the context of the biological diversity and to support them to develop ecologically oriented values.[25]
- Loving High School in Loving, New Mexico, publishes career and technical education opportunities for students. These include internship for students who are interested in science, STEM majors, or architecture. The school is making good connections with local businesses, which helps students get used to working in such environments.
- Lake View High School in Chicago, Illinois is the institution which offers early college credits for students. It trains students with majors such as STEM, humanities, music/ art, and languages.[26]
- Robert H. Smith School of Business offers select undergraduate students a year-round advanced course whereby students conduct financial analyses and security trades on Bloomberg Terminals to manage real investment dollars in the Lemma Senbet Fund.
In business education[edit]
As higher education continues to adapt to new expectations from students, experiential learning in business and accounting programs has become more important. For example, Clark & White (2010) point out that 'a quality university business education program must include an experiential learning component'.[27] With reference to this study, employers note that graduating students need to build skills in 'professionalism' – which can be taught via experiential learning. Students also value this learning as much as industry.
Learning styles also impact business education in the classroom. Kolb transposes four learning styles, Diverger, Assimilator, Accommodator and Converger, atop the Experiential Learning Model, using the four experiential learning stages to carve out 'four quadrants', one for each learning style. An individual’s dominant learning style can be identified by taking Kolb’s Learning Style Inventory (LSI). Robert Loo (2002) undertook a meta-analysis of 8 studies which revealed that Kolb’s learning styles were not equally distributed among business majors in the sample.[28] More specifically, results indicated that there appears to be a high proportion of assimilators and a lower proportion of accommodators than expected for business majors. Not surprisingly, within the accounting sub-sample there was a higher proportion of convergers and a lower proportion of accommodates. Similarly, in the finance sub-sample, a higher proportion of assimilators and lower proportion of divergers was apparent. Within the marketing sub-sample there was an equal distribution of styles. This would provide some evidence to suggest that while it is useful for educators to be aware of common learning styles within business and accounting programs, they should be encouraging students to use all four learning styles appropriately and students should use a wide range of learning methods.[28]
Professional education applications, also known as management training or organizational development, apply experiential learning techniques in training employees at all levels within the business and professional environment. Interactive, role-play based customer service training is often used in large retail chains.[29] Training board games simulating business and professional situations such as the Beer Distribution Game used to teach supply chain management, and the Friday Night at the ER game used to teach systems thinking, are used in business training efforts.[30]
In business[edit]
Experiential business learning is the process of learning and developing business skills through the medium of shared experience. The main point of difference between this and academic learning is more “real-life” experience for the recipient.[31][32][33]
This may include for example, learning gained from a network of business leaders sharing best practice, or individuals being mentored or coached by a person who has faced similar challenges and issues, or simply listening to an expert or thought leader in current business thinking.
Providers of this type of experiential business learning often include membership organisations who offer product offerings such as peer group learning, professional business networking, expert/speaker sessions, mentoring and/or coaching.
Comparisons[edit]
Experiential Learning is more efficient than passive learning such as reading or listening.[34]
Experiential learning is most easily compared with academic learning, the process of acquiring information through the study of a subject without the necessity for direct experience. While the dimensions of experiential learning are analysis, initiative, and immersion, the dimensions of academic learning are constructive learning and reproductive learning.[35] Though both methods aim at instilling new knowledge in the learner, academic learning does so through more abstract, classroom-based techniques, whereas experiential learning actively involves the learner in a concrete experience.
Benefits[edit]
- Experience real world: For example, students who major in Chemistry may have chances to interact with the chemical environment. Learners who have a desire to become businesspeople will have the opportunity to experience the manager position
- Opportunities for creativity: There is always more than one solution for a problem in the real world. Students will have a better chance to learn that lesson when they get to interact with real life experiences[36]
See also[edit]
Wikiversity has learning resources about Experiential learning |
People[edit]
Subjects[edit]
- Sudbury model of democratic education
References[edit]
- ^Felicia, Patrick (2011). Handbook of Research on Improving Learning and Motivation. p. 1003. ISBN1609604962.
- ^The Out of Eden Walk: An Experiential Learning Journey from the Virtual to the Real, Edutopia, January 3, 2014. Retrieved 2016-03-16
- ^Action Learning - How does it work in practice? MIT Sloan Management. Retrieved 2016-03-16[dead link]
- ^The Power of Experiential Learning, 4-H Cooperative Curriculum System. Retrieved 2016-03-16[dead link]
- ^Beard, Colin (2010). The Experiential Learning Toolkit: Blending Practice with Concepts. p. 20. ISBN9780749459345.
- ^ abItin, C. M. (1999). Reasserting the Philosophy of Experiential Education as a Vehicle for Change in the 21st Century. The Journal of Physical Education 22(2), p. 91-98.
- ^Breunig, Mary C. (2009). 'Teaching Dewey's Experience and Education Experientially'. In Stremba, Bob; Bisson, Christian A. (eds.). Teaching Adventure Education Theory: Best Practices. p. 122. ISBN9780736071260.
- ^Nicomachean Ethics, Book 2, Chase translation (1911).
- ^Dixon, Nancy M.; Adams, Doris E.; Cullins, Richard (1997). 'Learning Style'. Assessment, Development, and Measurement. p. 41. ISBN9781562860493.
- ^Hawtrey, Kim. 'Using Experiential Learning Techniques'.
- ^McCarthy, P. R., & McCarthy, H. M. (2006). When Case Studies Are Not Enough: Integrating Experiential Learning Into Business Curricula. Journal of Education for Business, 81(4), pp. 201-204.
- ^Kraft, R. G. (1994).Bike riding and the art of learning.In L. B. Barnes, C. Roland Christensen, & A. J. Hansen (Eds.), Teaching and the case method.Boston: Harvard Business School Press.
- ^Loo, R. (2002). A Meta-Analytic Examination of Kolb's Learning Style Preferences Among Business Majors. Journal of Education for Business, 77:5, 252-256
- ^ abcKolb, D. (1984). Experiential Learning: experience as the source of learning and development. Englewood Cliffs, NJ: Prentice Hall. p. 21
- ^http://www2.le.ac.uk/departments/gradschool/training/resources/teaching/theories/kolb Retrieved October 28, 2012.
- ^Merriam, S. B., Caffarella, R. S., & Baumgartner, L. M. (2007). Learning in adulthood: a comprehensive guide. San Francisco: John Wiley & Sons, Inc.
- ^ abcMoon, J. (2004). A Handbook of Reflective and Experiential Learning:Theory and Practice. London: Routledge Falmer. p. 126.
- ^Kolb, D (1984). Experiential Learning as the Science of Learning and Development. Englewood Cliffs, NJ: Prentice Hall.
- ^Chickering, A (1977). Experience and Learning. New York: Change Magazine Press. p. 63.
- ^Hutton, M. (1980). Learning from action: a conceptual framework, in S. Warner Weil and M. McGill (eds) Making Sense of Experiential Learning. Milton Keynes: SRHE/Open University Press. pp. 50–9, p. 51.
- ^Kompf, M., & Bond, R. (2001). Critical reflection in adult education. In T. Barer-Stein & M. Kompf (Eds.), The craft of teaching adults (p. 55). Toronto, ON: Irwin.
- ^ abcJacobson, M. & Ruddy, M. (2004) Open to outcome (p. 2). Oklahoma City, OK: Wood 'N' Barnes.
- ^Pfeiffer, W. & Jones, J. E. (1975). A Handbook of Structured Experiences for Human Relations Training. La Jolla, California: University Associates.
- ^Rodrigues, C. A. (2004). The importance level of ten teaching/learning techniques as rated by university business students and instructors. Journal of Management Development, 23(2), 169-182.
- ^ELENA project leader
- ^Staff, Noodle. '41 Most Innovative K–12 Schools in America'. Retrieved 2015-10-19.
- ^Clark, J., & White, G. (2010). 'Experiential Learning: A Definitive Edge In The Job Market'. American Journal of Business Education, 3(2), pp. 115-118.
- ^ abLoo, R. (2002). 'A Meta-Analytic Examination of Kolb's Learning Style Preferences Among Business Majors'. Journal of Education for Business, 77:5, 252-256
- ^https://multimediaplus.com/experiential-learning/
- ^Faria, Anthony J. '4'(PDF). Business Simulation Games after Thirty Years: Current Usage Levels in the United States in Gentry (ed.) Guide to Business Gaming and Experiential Learning. The University of Michigan: Nichols/GP Pub., 1990. pp. 36–47. ISBN978-0893973698. Retrieved 12 March 2014.
- ^'David A. Kolb on experiential learning'. infed.org. 2013-04-26. Retrieved 2018-10-15.
- ^Greenaway, Tim Pickles and Roger. 'Experiential learning articles + critiques of David Kolb's theory'. www.reviewing.co.uk. Retrieved 2018-10-15.
- ^Council, Young Entrepreneur. 'Seven Mentorship Methods And Opportunities Entrepreneurs Should Remember'. Forbes. Retrieved 2018-10-15.
- ^Skill Pyramid
- ^Stavenga de Jong, J. A., Wierstra, R. F. A. and Hermanussen, J. (2006) 'An exploration of the relationship between academic and experiential learning approaches in vocational education', British Journal of Educational Psychology. 76;1. pp. 155-169.
- ^'The Benefits of Experiential Learning'.
Retrieved from 'https://en.wikipedia.org/w/index.php?title=Experiential_learning&oldid=892738998'