Section 12.1
• What is machine learning?
• What are the strengths and limitations of traditional machine learning (as illustrated by ID3, for example)?
Section 12.2
• What is representation learning? How does it differ from traditional machine learning?
• How do LeCun, Bengio, and Hinton define deep learning?
• Why is the visual cortex an inspiration for deep learning theorists?
Section 12.3
• What does an autoencoder do? How does it work?
• What are the distinctive features of convolutional neural networks?
Section 12.4
• How does reinforcement differ from the types of learning that we have been considering up to now?
• Why are AlphaGo and AlphaGo Zero important?
• Is reinforcement learning important for thinking about the biological plausibility of deep learning?
Machine learning (video lecture by Mark Dredze, from videolectures.net)
Machine Learning: The Basics, with Ron Bekkerman (video from YouTube LinkedIn Tech Talks channel)
Deep Learning in the Visual Cortex, Part 1 (video lecture by Thomas Serre from Brown University)
How invariant feature selectivity is achieved in cortex (video lecture by Tatyana Sharpee from the Salk Institute of Biological Studies)
Introduction to Deep Reinforcement Learning (video lecture by Lex Fridman from MIT)
Deep Reinforcement Learning (video lecture from Stanford University)
12.1 Expert Systems and Machine Learning
Expert systems: Where are we? And where do we go from here? (paper by Randall Davis, 1982; in AI Magazine)
Expert systems: How far can they go? Part one (paper by Randall Davis, 1989; in AI Magazine)
Expert systems: How far can they go? Part two (paper by Randall Davis, 1989; in AI Magazine)
Does machine learning really work? (paper by Tom Mitchell, 1997; in AI Magazine)
12.2 Representation Learning and Deep Learning
Deep Learning (entry from Scholarpedia)
Models of visual cortex (entry from Scholarpedia)
Artificial intelligence (entry from Stanford Encyclopedia of Philosophy)
Deep learning (review paper by LeCun et al., 2015, in Nature)
Deep Learning: An MIT Press book (by Ian Goodfellow, Yoshua Bengio, and Aaron Courville, 2016)
Mastering the game of Go with deep neural networks and tree search (paper by Silver et al., 2016 , in Nature)
Selectivity and tolerance (“invariance”) both increase as visual information propagates from cortical area V4 to IT (paper by Rust and DiCarlo, 2010, in Journal of Neuroscience)
On invariance and selectivity in representation learning (paper by Anselmi, Rosasco, and Poggio, 2016, in journal Information and Inference)
Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition (paper by Cadieu et al., 2014, in PLOS Computational Biology)
ImageNet Classification with Deep Convolutional Neural Networks (conference paper by Krizhevsky, Sutskever, and Hinton, 2012, in Advances in neural information processing systems)
Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence (paper by Cichy et al., 2016, in Scientific Reports)
Large-Scale, High-Resolution Comparison of the Core Visual Object Recognition Behavior of Humans, Monkeys, and State-of-the-Art Deep Artificial Neural Networks (paper by Rajalingham et al., 2018, in Journal of Neuroscience)
Using neuroscience to develop artificial intelligence (paper by Ulman, 2020, in Science)
Building machines that learn and think like people (paper by Lake et al., 2017, in Behavioral and Brain Sciences)
12.3 The Machinery of Deep Learning
Tutorial on autoencoders (from Unsupervised Feature Learning and Deep Learning Tutorial)
Chapter on autoencoders from “Deep Learning: An MIT Press book” (by Goodfellow, Bengio, and Courville, 2016)
Chapter on ConvNets from “Deep Learning: An MIT Press book” (by Goodfellow, Bengio, and Courville, 2016)
Tutorial on ConvNets (an online tutorial from the Computer Science Department, Stanford University)
12.4 Deep Reinforcement Learning
Reinforcement Learning: An Introduction (2ed) (book by Sutton and Barto, 2018)
Dive into Deep Learning (deep learning self-taught materials)
An Introduction to Deep Reinforcement Learning (paper by François-Lavet et al., 2018, Foundations and Trends in Machine Learning)
A Framework for Mesencephalic Dopamine Systems Based on Predictive Hebbian Learning (paper by Montague, Dayan, and Sejnowski, 1996, in Journal of Neuroscience)
A Neural Substrate of Prediction and Reward (paper by Schultz et al., 1997, in Science)
Understanding dopamine and reinforcement learning: The dopamine reward prediction error hypothesis (paper by Glimcher, 2011, in Proceedings of the National Academy of Sciences)
Mastering the game of Go with deep neural networks and tree search (AlphaGo paper by Silver et al., 2016, in Nature)
Mastering the game of Go without human knowledge (AlphaGo Zero paper by Silver et al., 2017, in Nature)
