Skip to content
— CH. 1 · BIOLOGICAL INSPIRATION AND ORIGINS —

Convolutional neural network

~5 min read · Ch. 1 of 6
6 sections

  • In the 1950s and 1960s, neuroscientists David Hubel and Torsten Wiesel studied cat visual cortices to understand how brains process vision. They discovered that individual neurons in these animals respond only to stimuli within a restricted region of the visual field known as the receptive field. Neighboring cells possessed similar and overlapping receptive fields that covered the entire visual space systematically. Their 1968 paper identified two basic cell types: simple cells maximized output for straight edges with specific orientations, while complex cells had larger receptive fields insensitive to exact edge positions. This biological model inspired Kunihiko Fukushima to introduce a multilayer visual feature detection network in 1969 called the neocognitron. Fukushima's design featured elements where all units in one layer shared identical interconnecting coefficients across homogeneous arrangements. The weights were not trainable at this stage, but the architecture laid the essential core for future convolutional networks.

  • The neocognitron introduced in 1980 established two fundamental layer types: S-layers acting as shared-weight receptive-field layers and C-layers functioning as downsampling layers. In 1987, Toshiteru Homma, Les Atlas, and Robert Marks II presented a paper replacing multiplication with convolution in time during the first Conference on Neural Information Processing Systems. Alex Waibel and colleagues developed Time Delay Neural Networks in 1987 for phoneme recognition using gradient descent training methods. Yann LeCun and his team at AT&T Bell Laboratories published work in 1989 demonstrating backpropagation applied to handwritten ZIP code recognition. LeNet-5 emerged in 1995 as a pioneering seven-level network classifying hand-written numbers on checks digitized into 32 by 32 pixel images. This system integrated into NCR check reading systems and began fielding in American banks since June 1996. GPU implementations accelerated dramatically starting in 2004 when K.S. Oh and K. Jung showed standard neural networks could run twenty times faster on graphics processors. Dan Ciresan trained deep feedforward networks on GPUs in 2010 before extending this to CNNs in 2011 achieving sixtyfold acceleration compared to CPU training. AlexNet won the ImageNet Large Scale Visual Recognition Challenge in 2012 marking an early catalytic event for the AI boom.

  • A convolutional layer serves as the core building block where parameters consist of learnable filters or kernels with small receptive fields extending through full input volume depth. During forward passes each filter convolves across width and height computing dot products between entries producing two-dimensional activation maps. Stacking these activation maps along the depth dimension forms the complete output volume of the convolution layer. Local connectivity ensures neurons connect only to small regions of adjacent layers rather than all previous units. Parameter sharing schemes constrain neurons within single depth slices to use identical weights and biases across spatial positions. Pooling layers reduce spatial dimensions by combining outputs from neuron clusters into single units using max or average values. Max pooling selects maximum values from local clusters while average pooling calculates mean values across regions. A common form involves 2 by 2 filters applied with stride 2 subsampling every depth slice by half discarding seventy-five percent activations. Fully connected layers follow multiple convolution and pooling stages performing final classification via affine transformations involving matrix multiplication plus bias offsets.

  • Standard neural networks could run twenty times faster on graphics processing units compared to equivalent CPU implementations according to research published in 2004 by K.S. Oh and K. Jung. The first GPU implementation of a CNN described in 2006 by K. Chellapilla et al achieved fourfold speed improvements over standard CPU execution. Dan Ciresan trained deep feedforward networks on GPUs at IDSIA in 2010 before extending this approach to CNNs in 2011 achieving sixtyfold acceleration relative to CPU training speeds. These hardware advances enabled practical training of deeper networks previously impossible due to computational constraints. In 2011 the network won an image recognition contest achieving superhuman performance for the first time. Subsequent competitions saw state-of-the-art results on several benchmarks using these accelerated methods. AlexNet leveraged similar GPU-based architectures to win the ImageNet Large Scale Visual Recognition Challenge in 2012. This victory marked an early catalytic event driving widespread AI adoption across industries. Parallelization techniques available on Intel Xeon Phi processors offered alternative approaches though less attention was given to CPU optimization compared to GPU strategies.

  • CNNs process text images audio and video data enabling predictions from many different types of information sources. In 2012 error rates reached 0.23 percent on the MNIST database demonstrating rapid learning processes for image classification tasks. Facial recognition systems achieved large decreases in error rates with one paper reporting ninety-seven point six percent accuracy on five thousand six hundred still images. Video analysis treats space and time as equivalent dimensions performing convolutions simultaneously across both temporal and spatial domains. Natural language processing models achieve excellent results in semantic parsing search query retrieval sentence modeling and classification prediction. Ecological research uses CNNs to automatically detect animal behaviors including feeding social interactions and pose estimation from visual recordings. Drug discovery applications identify potential treatments by predicting molecule-biological protein interactions through three-dimensional chemical representations. The AtomNet system introduced in 2015 predicted novel candidate biomolecules for diseases like Ebola virus and multiple sclerosis. Checkers programs learned gameplay using co-evolution without prior human professional games achieving top rankings against expert players. Computer Go systems trained on human databases outperformed traditional search programs winning ninety-seven percent of games against GNU Go.

  • Dropout introduced in 2014 reduces overfitting by ignoring individual nodes during training stages with probability p while keeping others with probability 1-p. At testing time only a single network needs testing despite effectively generating exponential combinations during training. Stochastic pooling replaces deterministic operations with random selections according to multinomial distributions given activities within pooling regions. Artificial data generation techniques create new examples by cropping rotating or rescaling existing input images since mid-1990s. Early stopping halts learning before overfitting occurs though it limits the total learning process duration. Weight decay adds penalties proportional to sum or squared magnitude of weight vectors reducing acceptable model complexity levels. L2 regularization heavily penalizes peaky weight vectors preferring diffuse vectors encouraging networks to use all inputs slightly rather than some inputs extensively. Max norm constraints enforce absolute upper bounds on weight vector magnitudes typically order three to four through projected gradient descent methods. Data augmentation increases available training examples reducing overfitting degrees determined by both power and amount of received training. Transfer learning trains networks on larger datasets from related domains before fine-tuning weights using in-domain data for tiny training sets.

Up Next

Feedforward neural networkAlexNetDeep learning

Common questions

Who invented the neocognitron and when was it introduced?

Kunihiko Fukushima introduced the multilayer visual feature detection network called the neocognitron in 1969. This design featured elements where all units in one layer shared identical interconnecting coefficients across homogeneous arrangements.

When did LeNet-5 begin fielding in American banks?

LeNet-5 began fielding in American banks since June 1996 after emerging as a pioneering seven-level network classifying hand-written numbers on checks digitized into 32 by 32 pixel images in 1995.

What year did GPU implementations accelerate neural networks twenty times faster than CPUs?

Standard neural networks could run twenty times faster on graphics processors starting in 2004 when K.S. Oh and K. Jung showed this capability compared to equivalent CPU implementations.

How accurate were facial recognition systems on five thousand six hundred still images according to research?

Facial recognition systems achieved ninety-seven point six percent accuracy on five thousand six hundred still images while achieving large decreases in error rates with one paper reporting these results.

Which system predicted novel candidate biomolecules for diseases like Ebola virus and multiple sclerosis in 2015?

The AtomNet system introduced in 2015 predicted novel candidate biomolecules for diseases like Ebola virus and multiple sclerosis through three-dimensional chemical representations of molecule-biological protein interactions.

See all questions about Convolutional neural network →

All sources

168 references cited across the entry

  1. 1journalDeep learningYann LeCun et al. — 2015-05-28
  2. 2journalBackpropagation Applied to Handwritten Zip Code RecognitionY. LeCun et al. — December 1989
  3. 3bookConvolutional Neural Networks in Visual Computing: A Concise GuideRagav Venkatesan et al. — CRC Press — 2017-10-23
  4. 4bookRecent Trends and Advances in Artificial Intelligence and Internet of ThingsValentina E. Balas et al. — Springer Nature — 2019-11-19
  5. 5journalPowder-Bed Fusion Process Monitoring by Machine Vision With Hybrid Convolutional Neural NetworksYingjie Zhang et al. — September 2020
  6. 6journalResidue Number System-Based Solution for Reducing the Hardware Cost of a Convolutional Neural NetworkN.I. Chervyakov et al. — September 2020
  7. 7bookGuide to convolutional neural networks: a practical application to traffic-sign detection and classificationHamed Habibi Aghdam et al. — Springer — 2017-05-30
  8. 8journalApplication of the residue number system to reduce hardware costs of the convolutional neural network implementationM.V. Valueva et al. — Elsevier BV — 2020
  9. 9bookDeep content-based music recommendationAaron van den Oord et al. — Curran Associates, Inc. — 2013-01-01
  10. 10bookProceedings of the 25th international conference on Machine learning - ICML '08Ronan Collobert et al. — ACM — 2008-01-01
  11. 11book2020 42nd Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC)Oleksii Avilov et al. — IEEE — July 2020
  12. 12book2017 IEEE 19th Conference on Business Informatics (CBI)Avraam Tsantekidis et al. — IEEE — July 2017
  13. 13journalShift-invariant pattern recognition neural network and its optical architectureWei Zhang — 1988
  14. 14journalParallel distributed processing model with local space-invariant interconnections and its optical architectureWei Zhang — 1990
  15. 15bookArtificial Intelligence ResearchCoenraad Mouton et al. — Springer International Publishing — 2020
  16. 16journalHidden bias in the DUD-E dataset leads to misleading performance of deep learning in structure-based virtual screeningThomas Kurtzman — August 20, 2019
  17. 17journalNeocognitronK. Fukushima — 2007
  18. 18journalSubject independent facial expression recognition with robust face detection using a convolutional neural networkMasakazu Matusugu — 2003
  19. 19inlineConvolutional Neural Networks Demystified: A Matched Filtering Perspective Based Tutorial
  20. 20webConvolutional Neural Networks (LeNet) – DeepLearning 0.1 documentationLISA Lab
  21. 21arxivXception: Deep Learning with Depthwise Separable ConvolutionsFrançois Chollet — 2017-04-04
  22. 22webImageNet Classification with Deep Convolutional Neural NetworksAlex Krizhevsky
  23. 23conferenceA Neural Network for Speaker-Independent Isolated Word RecognitionKouichi Yamaguchi et al. — November 1990
  24. 24book2012 IEEE Conference on Computer Vision and Pattern RecognitionDan Ciresan et al. — Institute of Electrical and Electronics Engineers (IEEE) — June 2012
  25. 25arxivMulti-Scale Context Aggregation by Dilated ConvolutionsFisher Yu et al. — 2016-04-30
  26. 26arxivRethinking Atrous Convolution for Semantic Image SegmentationLiang-Chieh Chen et al. — 2017-12-05
  27. 27arxivContextual Convolutional Neural NetworksIonut Cosmin Duta et al. — 2021-08-16
  28. 28webLeNet-5, convolutional neural networksYann LeCun
  29. 29book2011 International Conference on Computer VisionMatthew D. Zeiler et al. — IEEE — November 2011
  30. 30citationA guide to convolution arithmetic for deep learningVincent Dumoulin et al. — 2018-01-11
  31. 31journalDeconvolution and Checkerboard ArtifactsAugustus Odena et al. — 2016-10-17
  32. 32journalComparing Object Recognition in Humans and Deep Convolutional Neural Networks—An Eye Tracking StudyLeonard Elia van Dyck et al. — 2021
  33. 33journalReceptive fields and functional architecture of monkey striate cortexD. H. Hubel et al. — 1968-03-01
  34. 34bookBrain and visual perception: the story of a 25-year collaborationDavid H. Hubel and Torsten N. Wiesel — Oxford University Press US — 2005
  35. 35journalReceptive fields of single neurones in the cat's striate cortexDH Hubel et al. — October 1959
  36. 36journalVisual feature extraction by a multilayered network of analog threshold elementsK. Fukushima — 1969
  37. 37arxivAnnotated History of Modern AI and Deep LearningJuergen Schmidhuber — 2022
  38. 38journalNeocognitron: A Self-organizing Neural Network Model for a Mechanism of Pattern Recognition Unaffected by Shift in PositionKunihiko Fukushima — 1980
  39. 39arxivSearching for Activation FunctionsPrajit Ramachandran et al. — October 16, 2017
  40. 40journalAn Artificial Neural Network for Spatio-Temporal Bipolar Patterns: Application to Phoneme ClassificationToshiteru Homma — 1987
  41. 41conferencePhoneme Recognition Using Time-Delay Neural NetworksAlex Waibel — 18 December 1987
  42. 42inlineAlexander Waibel et al., Phoneme Recognition Using Time-Delay Neural Networks IEEE Transactions on Acoustics, Speech, and Signal Processing, Volume 37, No. 3, pp. 328. - 339 March 1989.
  43. 43encyclopediaConvolutional networks for images, speech, and time seriesYann LeCun et al. — The MIT press — 1995
  44. 44inlineJohn B. Hampshire and Alexander Waibel, Connectionist Architectures for Multi-Speaker Phoneme Recognition , Advances in Neural Information Processing Systems, 1990, Morgan Kaufmann.
  45. 45conferenceA Study on Data Augmentation of Reverberant Speech for Robust Speech RecognitionTom Ko et al. — March 2018
  46. 46inlineDenker, J S, Gardner, W R, Graf, H. P, Henderson, D, Howard, R E, Hubbard, W, Jackel, L D, BaIrd, H S, and Guyon (1989) Neural network recognizer for hand-written zip code digits , AT&T Bell Laborator
  47. 47inlineY. LeCun, B. Boser, J. S. Denker, D. Henderson, R. E. Howard, W. Hubbard, L. D. Jackel, Backpropagation Applied to Handwritten Zip Code Recognition ; AT&T Bell Laboratories
  48. 48book1993 (4th) International Conference on Computer VisionJ Weng et al. — IEEE — 1993
  49. 49journalDeep LearningJürgen Schmidhuber — 2015
  50. 50bookLearning algorithms for classification: A comparison on handwritten digit recognitionLecun, Y. et al. — World Scientific — August 1995
  51. 51journalGradient-based learning applied to document recognitionY. Lecun et al. — November 1998
  52. 52journalError Back Propagation with Minimum-Entropy Weights: A Technique for Better Generalization of 2-D Shift-Invariant NNsWei Zhang — 1991
  53. 53journalImage processing of human corneal endothelium based on a learning networkWei Zhang — 1991
  54. 54journalComputerized detection of clustered microcalcifications in digital mammograms using a shift-invariant artificial neural networkWei Zhang — 1994
  55. 55inlineDaniel Graupe, Ruey Wen Liu, George S Moschytz."Applications of neural networks to medical signal processing ". In Proc. 27th IEEE Decision and Control Conf., pp. 343–347, 1988.
  56. 56inlineDaniel Graupe, Boris Vern, G. Gruener, Aaron Field, and Qiu Huang. "Decomposition of surface EMG signals into single fiber action potentials by means of neural network ". Proc. IEEE International Symp
  57. 57inlineQiu Huang, Daniel Graupe, Yi Fang Huang, Ruey Wen Liu."Identification of firing patterns of neuronal signals." In Proc. 28th IEEE Decision and Control Conf., pp. 266–271, 1989.
  58. 58journalGPU implementation of neural networks.KS Oh et al. — 2004
  59. 59conference12th International Conference on Document Analysis and Recognition (ICDAR 2005)Dave Steinkraus et al. — 2005
  60. 60bookTenth International Workshop on Frontiers in Handwriting RecognitionKumar Chellapilla et al. — Suvisoft — 2006
  61. 61journalA fast learning algorithm for deep belief nets.GE Hinton et al. — Jul 2006
  62. 62journalGreedy Layer-Wise Training of Deep NetworksYoshua Bengio et al. — 2007
  63. 63journalEfficient Learning of Sparse Representations with an Energy-Based ModelMarcAurelio Ranzato et al. — 2007
  64. 64bookProceedings of the 26th Annual International Conference on Machine LearningR Raina et al. — ICML '09: Proceedings of the 26th Annual International Conference on Machine Learning — 14 June 2009
  65. 65journalDeep big simple neural nets for handwritten digit recognition.Dan Ciresan et al. — 2010
  66. 66journalFlexible, High Performance Convolutional Neural Networks for Image ClassificationDan Ciresan — 2011
  67. 67webIJCNN 2011 Competition result table2010
  68. 68webHistory of computer vision contests won by deep CNNs on GPUJürgen Schmidhuber — 17 March 2017
  69. 69journalCHAOS: a parallelization scheme for training convolutional neural networks on Intel Xeon PhiAndre Viebke et al. — 2019
  70. 70conference2015 IEEE 17th International Conference on High Performance Computing and Communications, 2015 IEEE 7th International Symposium on Cyberspace Safety and Security, and 2015 IEEE 12th International Conference on Embedded Software and SystemsAndre Viebke et al. — IEEE 2015 — 2015
  71. 71journalImageNet Classification with Deep Convolutional Neural NetworksGeoffrey Hinton — 2012
  72. 72bookHands-on Machine Learning with Scikit-Learn, Keras, and TensorFlowAurélien Géron — O'Reilly Media — 2019
  73. 73journalA Survey of Convolutional Neural Networks: Analysis, Applications, and ProspectsZewen Li et al. — December 2022
  74. 74webCS231n Convolutional Neural Networks for Visual Recognition
  75. 75journalPooling in convolutional neural networks for medical image analysis: a survey and an empirical studyRajendran Nirthika et al. — 2022-04-01
  76. 76journalWhy do deep convolutional networks generalize so poorly to small image transformations?Aharon Azulay et al. — 2019
  77. 77conferenceEvaluation of Pooling Operations in Convolutional Architectures for Object RecognitionDominik Scherer et al. — Springer — 2010
  78. 78arxivFractional Max-PoolingBenjamin Graham — 2014-12-18
  79. 79arxivStriving for Simplicity: The All Convolutional NetJost Tobias Springenberg et al. — 2014-12-21
  80. 80journalFine-Grained Vehicle Classification With Channel Max Pooling Modified CNNsZhanyu Ma et al. — Institute of Electrical and Electronics Engineers (IEEE) — 2019
  81. 81journalA Comparison of Pooling Methods for Convolutional Neural NetworksAfia Zafar et al. — 2022-08-29
  82. 82citationPooling Methods in Deep Neural Networks, a ReviewHossein Gholamalinezhad et al. — 2020-09-16
  83. 83journalA theory of steady-state activity in nerve-fiber networks: I. Definitions and preliminary lemmasAlston S. Householder — June 1941
  84. 84journalImageNet classification with deep convolutional neural networksAlex Krizhevsky et al. — 2017-05-24
  85. 85journalAppropriate number and allocation of ReLUs in convolutional neural networksVadim Romanuke — 2017
  86. 86conferenceDeep sparse rectifier neural networksXavier Glorot et al. — 2011
  87. 87journalImagenet classification with deep convolutional neural networksA. Krizhevsky — 2012
  88. 88bookICASSP 2021 - 2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)Antonio H. Ribeiro et al. — 2021
  89. 89bookArtificial Intelligence ResearchJohannes C. Myburgh et al. — Springer International Publishing — 2020
  90. 90bookMaking Convolutional Networks Shift-Invariant AgainZhang Richard — 2019-04-25
  91. 91journalSpatial Transformer NetworksMax Jadeberg et al. — 2015
  92. 92bookDynamic Routing Between CapsulesSara Sabour et al. — 2017-10-26
  93. 93journalInductive conformal predictor for convolutional neural networks: Applications to active learning for image classificationSergio Matiz et al. — 2019-06-01
  94. 94journalDeep Learning With Conformal Prediction for Hierarchical Analysis of Large-Scale Whole-Slide Tissue ImagesHåkan Wieslander et al. — February 2021
  95. 95journalDropout: A Simple Way to Prevent Neural Networks from overfittingNitish Srivastava — 2014
  96. 96journalRegularization of Neural Networks using DropConnect ICML 2013 JMLR W&CP2013-02-13
  97. 97arxivStochastic Pooling for Regularization of Deep Convolutional Neural NetworksMatthew D. Zeiler et al. — 2013-01-15
  98. 98journalBest Practices for Convolutional Neural Networks Applied to Visual Document Analysis – Microsoft ResearchJohn Platt et al. — August 2003
  99. 99arxivImproving neural networks by preventing co-adaptation of feature detectorsGeoffrey E. Hinton et al. — 2012
  100. 100webDropout: A Simple Way to Prevent Neural Networks from Overfitting
  101. 101journalSome demonstrations of the effects of structural descriptions in mental imageryGeoffrey Hinton — 1979
  102. 102inlineJ. Hinton, Coursera lectures on Neural Networks, 2012, Url:
  103. 103webThe inside story of how AI got good enough to dominate Silicon ValleyDave Gershgorn — 18 June 2018
  104. 104journalFace Recognition: A Convolutional Neural Network ApproachSteve Lawrence — 1997
  105. 105journalA Convolutional Neural Network Approach for Objective Video Quality AssessmentPatrick Le Callet — 2006
  106. 106webImageNet Large Scale Visual Recognition Competition 2014 (ILSVRC2014)
  107. 107conferenceIEEE Conference on Computer Vision and Pattern Recognition, CVPR 2015, Boston, MA, USA, June 7–12, 2015Christian Szegedy et al. — IEEE Computer Society — 2015
  108. 108arxivImage Net Large Scale Visual Recognition ChallengeOlga Russakovsky et al. — 2014
  109. 109newsThe Face Detection Algorithm Set To Revolutionize Image SearchFebruary 16, 2015
  110. 110bookHuman Behavior UnterstandingMoez Baccouche et al. — Springer Berlin Heidelberg — 2011-11-16
  111. 111journal3D Convolutional Neural Networks for Human Action RecognitionShuiwang Ji et al. — 2013-01-01
  112. 112arxivVideo-based Sign Language Recognition without Temporal SegmentationJie Huang et al. — 2018
  113. 113inlineKarpathy, Andrej, et al. "Large-scale video classification with convolutional neural networks ." IEEE Conference on Computer Vision and Pattern Recognition (CVPR). 2014.
  114. 114arxivTwo-Stream Convolutional Networks for Action Recognition in VideosKaren Simonyan et al. — 2014
  115. 115journalSegment-Tube: Spatio-Temporal Action Localization in Untrimmed Videos with Per-Frame SegmentationLe Wang et al. — 2018-05-22
  116. 116conference2018 25th IEEE International Conference on Image Processing (ICIP)Xuhuan Duan et al. — 25th IEEE International Conference on Image Processing (ICIP) — 2018
  117. 117conferenceConvolutional Learning of Spatio-temporal FeaturesGraham W. Taylor et al. — Springer-Verlag — 2010-01-01
  118. 118bookCVPR 2011Q. V. Le et al. — IEEE Computer Society — 2011-01-01
  119. 119arxivA Deep Architecture for Semantic ParsingEdward Grefenstette et al. — 2014-04-29
  120. 120journalLearning Semantic Representations Using Convolutional Neural Networks for Web Search – Microsoft ResearchGregoire Mesnil et al. — April 2014
  121. 121arxivA Convolutional Neural Network for Modelling SentencesNal Kalchbrenner et al. — 2014-04-08
  122. 122arxivConvolutional Neural Networks for Sentence ClassificationYoon Kim — 2014-08-25
  123. 123inlineCollobert, Ronan, and Jason Weston. "A unified architecture for natural language processing: Deep neural networks with multitask learning ."Proceedings of the 25th international conference on Machine
  124. 124arxivNatural Language Processing (almost) from ScratchRonan Collobert et al. — 2011-03-02
  125. 125arxivComparative study of CNN and RNN for natural language processingW Yin et al. — 2017-03-02
  126. 126arxivAn empirical evaluation of generic convolutional and recurrent networks for sequence modelingS. Bai et al. — 2018
  127. 127journalDetecting dynamics of action in text with a recurrent neural networkN. Gruber — 2021
  128. 128journalApproximation Theory of Convolutional Architectures for Time Series ModellingJ. Haotian et al. — 2021
  129. 129journalDeepEthogram, a machine learning pipeline for supervised behavior classification from raw pixelsJames P Bohnslav et al. — 2021-09-02
  130. 130journalAutomated monitoring of honey bees with barcodes and artificial intelligence reveals two distinct social networks from a single affiliative behaviorTim Gernat et al. — 2023-01-27
  131. 131journalAutomatically identifying, counting, and describing wild animals in camera-trap images with deep learningMohammad Sadegh Norouzzadeh et al. — 2018-06-19
  132. 132biorxivA General Method for Detection and Segmentation of Terrestrial Arthropods in ImagesAsger Svenning et al. — 2025-04-14
  133. 133biorxivNew idtracker.ai: rethinking multi-animal tracking as a representation learning problem to increase accuracy and reduce tracking timesJordi Torrents et al. — 2025-06-02
  134. 134journalDeepLabCut: markerless pose estimation of user-defined body parts with deep learningAlexander Mathis et al. — September 2018
  135. 135journalDeepPoseKit, a software toolkit for fast and robust animal pose estimation using deep learningJacob M Graving et al. — 2019-10-01
  136. 136journalPublisher Correction: SLEAP: A deep learning system for multi-animal pose trackingTalmo D. Pereira et al. — May 2022
  137. 137journalDeepBehavior: A Deep Learning Toolbox for Automated Analysis of Animal and Human Behavior Imaging DataAhmet Arac et al. — 2019-05-07
  138. 138conferenceTime-Series Anomaly Detection Service at Microsoft Proceedings of the 25th ACM SIGKDD International Conference on Knowledge Discovery & Data MiningHansheng Ren et al. — 2019
  139. 139arxivAtomNet: A Deep Convolutional Neural Network for Bioactivity Prediction in Structure-based Drug DiscoveryIzhar Wallach et al. — 2015-10-09
  140. 140arxivUnderstanding Neural Networks Through Deep VisualizationJason Yosinski et al. — 2015-06-22
  141. 141newsToronto startup has a faster way to discover effective medicines
  142. 142webStartup Harnesses Supercomputers to Seek Cures2015-05-27
  143. 143journalEvolving neural networks to play checkers without relying on expert knowledgeK Chellapilla et al. — 1999
  144. 144journalEvolving an expert checkers playing program without using human expertiseK. Chellapilla et al. — 2001
  145. 145bookBlondie24: Playing at the Edge of AIDavid Fogel — Morgan Kaufmann — 2001
  146. 146arxivTeaching Deep Convolutional Neural Networks to Play GoChristopher Clark et al. — 2014
  147. 147arxivMove Evaluation in Go Using Deep Convolutional Neural NetworksChris J. Maddison et al. — 2014
  148. 148webAlphaGo – Google DeepMind
  149. 149arxivAn Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence ModelingShaojie Bai et al. — 2018-04-19
  150. 150arxivConditional Time Series Forecasting with Convolutional Neural NetworksAnastasia Borovykh et al. — 2018-09-17
  151. 151arxivTime-series modeling with undecimated fully convolutional neural networksRoni Mittelman — 2015-08-03
  152. 152arxivProbabilistic Forecasting with Temporal Convolutional Neural NetworkYitian Chen et al. — 2019-06-11
  153. 153journalConvolutional neural networks for time series classiBendong Zhao et al. — 2017-02-01
  154. 154arxivQCNN: Quantile Convolutional Neural NetworkGábor Petneházi — 2019-08-21
  155. 155inlineDurjoy Sen Maitra; Ujjwal Bhattacharya; S.K. Parui, "CNN based common approach to handwritten character recognition of multiple scripts" , in Document Analysis and Recognition (ICDAR), 2015 13th Inter
  156. 156webNIPS 20172017-10-20
  157. 157bookArtificial Intelligence Applications and InnovationsJinliang Zang et al. — Springer International Publishing — 2018
  158. 158journalAction Recognition by an Attention-Aware Temporal Weighted Convolutional Neural NetworkLe Wang et al. — 2018-06-21
  159. 159arxivDistributed Deep Q-LearningHao Yi Ong et al. — 2015-08-18
  160. 160journalHuman-level control through deep reinforcement learningVolodymyr Mnih — 2015
  161. 161journalSelf-segmentation of sequences: automatic formation of hierarchies of sequential behaviorsR. Sun et al. — June 2000
  162. 162webConvolutional Deep Belief Networks on CIFAR-10
  163. 163bookProceedings of the 26th Annual International Conference on Machine LearningHonglak Lee et al. — ACM — 1 January 2009
  164. 164bookHierarchical Neural Networks for Image InterpretationSven Behnke — Springer — 2003
  165. 165newsIntroduction to Machine Learning, Neural Networks, and Deep LearningRene Y. Choi et al. — February 2020
  166. 166citationBreaking the Code on Broken Tablets: The Learning Challenge for Annotated Cuneiform Script in Normalized 2D and 3D Datasets2019
  167. 167citationHeiCuBeDa Hilprecht – Heidelberg Cuneiform Benchmark Dataset for the Hilprecht CollectionheiDATA – institutional repository for research data of Heidelberg University — 2019-06-07
  168. 168citationPeriod Classification of 3D Cuneiform Tablets with Geometric Neural NetworksBartosz Bogacz et al. — 2020