15-386/686 Neural Computation

Week 1 and 2 (1/18, 1/25): The Computer and the Brain

• Here are some classics on computational neuroscience and neurophysiology that might be of interest -- we might read some of them during the semester. Week 1: Mert Inan will present the classic Perceptron paper by Rosenblatt. Dr. Wenhao Zhang will present his theoretical research relating co-variaiblity of neurons and Bayesian inference. Week 2: Martin Ma will present McCulloch and Pitt Logical Calculus paper. Vincent Mai will present Lettvin, Maturana, McCulloch and Pitt's frog's eye paper, for modern reference, check out Gollisch and Meister's paper on retina.
• Marr, D. (1982) Chapter 1: The Philosophy and the Approach Vision 8-38 .
• Lettvin, Maturana, McCulloch and Pitt. (1959) What the frog's eye tells the frog's brain Proceedings of the IRE 1940-1959 .
• John von Neumann (1958) the Computer and the Brain Stillman Lectures 1-80 .
• Horace Barlow (1961) Possible Principles. Underlying the Transformations of Sensory Messages Sensory Communication, Ed W.Rosenblith Ch13 , pp. 217-234.
• McCulloch and Pitts (1943) A Logical Calculus of Ideas Immanent in Nervous Activity Bulletin of Mathematical Biophysics 5: 115-133.
• Norbert Wiener (1948) Cybernetics: Control and Communication in the Animal and the Machine. MIT Press 1-203.
• Frank Rosenblatt (1958) The Perceptron: A Probabilistic model for information storage and organization in the brain. Psychological Review 65:6. 386-408.

Week 3 (2/1): Relating Structures, Circuits and Functions -- Retina Example

• Retina might be the best understood part of the brain in terms of detailed circuits and functions. We will read a couple papers on the retina by Markus Meister's half century after Lettvin et al's paper on frog's eyes. Julie (Geyang) will present the 2010 Neuron review paper. Rockwell Hall will present the (2017) Current biology paper.
• Real, Asari, Gollisch and Meister (2017) Neural Circuit Inference from Function to Structure Current Biology 27(2): 189-198.
• Gollisch and Meister (2010) Eye Smarter than Scientists Believed: Neural Computations in Circuits of the Retina Neuron 65: 151-164.
• Seung and Sumbul (2014) Neuronal Cell Types and Connectivity: Lessons from the Retina Neuron 83(6): 1262-1272.

Week 4 (2/8): Synaptic Plasticity and Biological Learning Rules

• Rules for synaptic plasticity and learning are diverse and complex. This week or maybe next, we will investigate the state of the knowledges on neuronal learning rules. Why are there different synaptic plasticities? What learning rules they are actually implementing. We will choose a couple papers from among the following to begin our inquiry. Week 4: Jinke Liu will read Abbott's paper Taming the beast. Franlin Ruogu Lin will read Sejnowski's paper on Neural Computational theory of plasticity. Week 5. Mert Innan will read papers on BCM rule: Lim et al. (Brunel) (2015) Nature Neuroscience, and for background part of the original BCM paper (1982). Martin Ma will read Minden, Pehlevan and Chklovskii's (2018) paper on on-line PCA.
• Abbott and Nelson (2000) Snaptic plasticity, taming the beast. Nature Neuroscience Supplement Review (3): 1178-1`183.
• Gina Turrigiano (2008) The self-tuning neurons: synaptic scaling of excitatory synapses Cell (135) 422-435.
• Markram, Lubke, Frotscher, Sakmann (1997) Regulation of Synaptic Efficacy by Coincidence of Postsynaptic APs and EPSPs Science (275) 213-215.
• Bi and Poo (1998) Synaptic Modifications in Cultured Hippocampal Neurons: Dependence on Spike Timing, Synaptic Strength, and Postsynaptic Cell Type J. Neuroscience 18(24) 10464-10472.
• Suvarathan (2018) Beyond STDP -- towards diverse and functionally relevant plasticity rules Current Opinion in Neurobiology (54)12-19.
• Finnie, Bear and Cooke (2017) Plasticity and Memory in Cerebral Cortex Learning and memory: A Comprehensive Reference (3) 233-261.
• Moldakarimov andSejnowski (2017) Neural Computation Theories of Learning Learning and memory: A Comprehensive Reference (3) 579-589.

Week 5 (2/15): More on Biological Learning Rules (BCM and PCA)

• We will continue our investigation on biologically learning rules. Mert Innan will read papers on BCM rule: Lim et al. (Brunel) (2015) Nature Neuroscience, and for background part of the original BCM paper (1982). Martin Ma will read Minden, Pehlevan and Chklovskii's (2018) paper on on-line PCA.
• Oja (1982) A simplified neuron modewl as a principal component analyzer Journal of Mathematical Biology 15: 267-273.
• Sanger (1989) Optimal unsupewrvised learning in a single-layer linear feedforward neural network. Neural Networks 2: 459-473.
• Minden, Pehlevan and Chklovskii (2018) Biologically Plausible Online Principal Component Analysis Without Recurrent Neural Dynamics arXiv 1810.06966v2
• Bienenstock, Cooper and Munro (1982) Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex J. Neuroscience 2(1): 32-48.
• Lim, ... Brunel (2015) Inferring learning rules from distribution of firing rates in cortical neurons Nature Neuroscience 18(12): 1804-1810.

Week 6 and 7 (2/22, 3/1): Sparse Coding and its Implication in Neural Networks

• We will begin to explore the functional and computaitonal benefits of sparse coding and extreme sparse coding in the context of associative memory, recurrent networks and deep networks. For the first week, Lewis Li will read HT Kung's paper on Deep Sparse-coded Network, and Sam Liu will read Sparse distributed memory of Kanerva. For the second week, Franklin Lin will present Babdi and Sompolinsky (2014), and Vicent Ma, assisted by Jinke Liu, will present Zemel and Urtasun's Parsimonious networks.
• Olshausen and Field (2004) Current Opinions in Biology 4(481-487)
• Tang et al. (2018) Large-scale two-photon imaging revealed super-sparse population codes in the V1 superficial layer of awake monkeys e-life eLife 2018;7:e33370
• Y. Gwan, M. Cha, HT Kung (2018) Deep Sparse-coded Network (DSN) Archived 1-6
• Pentti Kanerva (1993) Sparse Distributed Memory and Related Models. In M.H. Hassoun, ed., Associative Neural Memories: Theory and Implementation, Oxford University Press 50-76
• G. Palm (2013): Neural Associative Memories and Sparse coding. Neural Networks 37(165-171)
• Liao, R, Schwing, A, Zemel, R, Urtasun R, Learning deep parsimonious representations. NIPS 2016
• Babadi and Sompolinsky (2014) Sparseness and Expansion in Sensory Representations Neuron 83: 1213-1226.

Week 8 (3/8): Neural Codes -- Variability and Covariability

• The variability of the neural codes have been proposed to represent, in the form of population activities, probability distributions and uncertainty, and priors. The covariability of neural responses have been related to statistical priors used in inference. This week, we will start reading some important papers to better understand these issues. Summer will read Orban et al. (2016), and Mert will read Okun et al. (2015).
• Ma, Beck, Latham and Pouget (2006) Bayesian inference with probabilistic population codes. Nature Neuroscience 9(11): 1432-1483.
• Fiser, Berkes, Orban and Lengyel (2010) Statistically optimal perception and learning: from behavior to neural representations. Trends in Cognitive Science 14(3): 119-130.
• Shivkumar, Lange, Chattoraj, Haefner (2018) A probabilistic population code based on neural samples. NIPS 2018
• Ganguli and Simoncelli (2014) Efficient sensory encoding and Bayesian inference with hterogeneous neural populations. Neural Computation 26: 2103-2134.
• Lin, Okun, Caradini and Harris (2005) The Nature of Shared Cortical Variability Neuron 87:644-656.
• Okun,...., Harris (2015) Diverse coupling of neurons to populations in sensory cortex Nature 251: 511-515
• Orban, Bekes, Fiser and Lengyel (2016) Neural Variability and Sampling-Based Probabilistic Representations in the Visual Cortex Neuron 92: 530-543.

Week 9 (3/18): Attractors and Memories

• This week, we are studying attractors for associative memories, Hopfield nets and Boltzmann machines. Our journal club will read some papers that cover other aspects of memories. First, we will look into theories of working memories. Simin Li will present papers by Alan Baddeley and colleagues on working memory and phonological loops, which Manuel Blum and Lenore Blum have drawn heavily on to develop their conscious turing machine (CTM) to model consciouness in human, animals and machine.
• Baddeley, Gathercole and Papagno (1998) The phonological loop as a language learning device. Psychological Review Vol. 105. No. 1, 158-173.
• Baddeley (1997) The role of memory in cognition: working memory. Memory and Cognition(?) Chapter 4

Week 10 (3/25-4/5): Attractors and Prototype Networks

• This week, we will study some recent work on the integration of assocation memories and deep convolution neural networks, called Prototype networks. Julie Zhang will present Zemel's paper. Sam Liu will present Si Wu's papers. Unfortunately, Sam Liu was sick. We will revisit Si Wu's works later.
• Snell, Swersky and Zemel (2017) Prototypical Networks for Few-shot Learning. NIPS 2007.
• Ji, Zou, Li, Huang, Mi, Wu (2018) Neural Information Processing in Hierarchical Prototypical Networks ICON 2018
• Robust Classification with Convolutional Prototype Learning CVPR 2018
• Fukushima (1984) A hierarchical neural network model for associative memory. Biol Cybern. 50(2):105-13.
• Wu (2019) Push‐pull Feedback Implements Hierarchical Associative Memory In Preparation 2019
• Zou, Ji, Liu, Huang, Mi, Wang, Wu. (2019) Learning, Storing, and Disentangling Correlated Patterns in Neural Networks In Preparation 2019

Week 11 (4/13): (Spring Carnival) Recurrent Neural Networks and Target Propagation

• This week, we have a optinal and shorter journal club, to be followed by Lee lab meeting. For the journal club Zikai Li will present Sussillo's paper on the capacity of recurrent neural networks. Mert Inan will present Richard Blake's works on dendritic computation for implementing Backpropagation or target propoagation (actually done last week in place of Sam's presentation).
• Collins, Sohl-Dickstein and Sussillo (2017) Capacity and trainability in recurrent neural networks ICLR 2017
• Lee, Zhang, Fischer and Bengio (2015) Difference Target Propagation Machine Learning and Knowledge Discovery in Databases. ECML PKDD 2015. Lecture Notes in Computer Science, vol 9284. Springer
• Guerguiev, Lillicrap, Richards (2017) Towards deep learning with segregated dendrites e-life 2017;6:e22901 DOI: https://doi.org/10.7554/eLife.22901
• Bartunov et al. Lillicrap (2017) Assessing the Scalability of Biologically-Motivated Deep Learning Algorithms and Architectures NIPS 2018

Week 12 (4/19): Generative Models and Transformer Networks

• This week, we will explore generative models and a couple interesting neural models of neural networks called transformer and attention networks. Hal will give a presentation of George's paper on breaking CAPTCHAs, and Martin Ma will present papers related to transformer networks for natural language processing.
• George, D et al. (2017) A generative vision model that trains with high data efficiency and breaks text-based CAPTCHAs. Science 358, 1271.
• Vaswani et al. (2017) Attention is all you need. NIPS 31th conference, Long Beach, CA.
• Parmar et al. (2018) Image Transformer. ICML Proceedgins for Internationa Conference on Machine Learning.
• Devlin et al. (2018) BERT: Pre-training of deep bidirectional transformers for language understanding. arXiv:1810.04805

Week: Imagination and Introspection Network

• This week, we will explore models that exploit the role of imagination in deep learning.
• Tu (2017) Learning generative models via discriminnative approahces. CVPR 2007.
• Jin et al. (2018( Introspective classification with convolutional nets NIPS 2017.
• Lazarow et al. (2017) Introspective Neural Networks for generative modeling. ICCV 2017.
• Weber et al. (2017) Imagination-Augmented Agents for deep reinforcement learning. NIPS 2017.
• Wang et al. (2017) Learning Robust Object Recognition Using Composed Scenes from Generative Models 14th Conference of Computer Vision and Robotics. 2017.
• Pascanu et al. (2017) Learning model-based planning from scratch. ? 2017.

Week: Reinforcement Learning, Curiosity Learning and Song Birds

• This week, we will focus on reinforcement learning, and a particular interesting case of song-bird learning.
• Burda et al. .. Efros (2018) Large-Scale Study of Curiosity-Driven Learning ICLR 2018
• Pathak, Agrawal, Efros and Darrell (2017) Curiosity-driven Exploration by Self-supervised Prediction. CVPR 2017
• Fee and Goldberg (2011) A hypothesis for basal ganglia-dependent reinforcement learning in the songbird. Neuroscience 198: 152-170.
• Gadagkar et al. (2016) Dopamine neurons encode performance error in singing birds Science 354: 1278-1282.
• Mackevicius and Fee (2018) Building a state space for song learning Current Opinion in Neurobiology 49: 59-68.