A deep learning framework for neuroscience

Systems neuroscience seeks explanations for how the brain implements a wide variety of perceptual, cognitive and motor tasks. Conversely, artificial intelligence attempts to design computational systems based on the tasks they will have to solve. In artificial neural networks, the three components specified by design are the objective functions, the learning rules and the architectures. With the growing success of deep learning, which utilizes brain-inspired architectures, these three designed components have increasingly become central to how we model, engineer and optimize complex artificial learning systems. Here we argue that a greater focus on these components would also benefit systems neuroscience. We give examples of how this optimization-based framework can drive theoretical and experimental progress in neuroscience. We contend that this principled perspective on systems neuroscience will help to generate more rapid progress. A deep network is best understood in terms of components used to design it—objective functions, architecture and learning rules—rather than unit-by-unit computation. Richards et al. argue that this inspires fruitful approaches to systems neuroscience.

[1]  Nikolaus Kriegeskorte,et al.  Cognitive computational neuroscience , 2018, Nature Neuroscience.

[2]  Demis Hassabis,et al.  Mastering the game of Go without human knowledge , 2017, Nature.

[3]  David J. C. MacKay,et al.  Information Theory, Inference, and Learning Algorithms , 2004, IEEE Transactions on Information Theory.

[4]  Peter Dayan,et al.  A Neural Substrate of Prediction and Reward , 1997, Science.

[5]  Razvan Pascanu,et al.  Vector-based navigation using grid-like representations in artificial agents , 2018, Nature.

[6]  Craig M. Vineyard,et al.  Crossing the Cleft: Communication Challenges Between Neuroscience and Artificial Intelligence , 2020, Frontiers in Computational Neuroscience.

[7]  Justus M. Kebschull,et al.  High-Throughput Mapping of Single-Neuron Projections by Sequencing of Barcoded RNA , 2016, Neuron.

[8]  Erkki Oja,et al.  Simple Neuron Models for Independent Component Analysis , 1996, Int. J. Neural Syst..

[9]  Michael Robert DeWeese,et al.  A Sparse Coding Model with Synaptically Local Plasticity and Spiking Neurons Can Account for the Diverse Shapes of V1 Simple Cell Receptive Fields , 2011, PLoS Comput. Biol..

[10]  Surya Ganguli,et al.  A Unified Theory Of Early Visual Representations From Retina To Cortex Through Anatomically Constrained Deep CNNs , 2019, bioRxiv.

[11]  Filip De Turck,et al.  VIME: Variational Information Maximizing Exploration , 2016, NIPS.

[12]  Joel Z. Leibo,et al.  Prefrontal cortex as a meta-reinforcement learning system , 2018, bioRxiv.

[13]  G. Fishell,et al.  Interneuron cell types are fit to function , 2014, Nature.

[14]  Anitha Pasupathy,et al.  'Artiphysiology' reveals V4-like shape tuning in a deep network trained for image classification , 2018, eLife.

[15]  Nathan Intrator,et al.  Objective function formulation of the BCM theory of visual cortical plasticity: Statistical connections, stability conditions , 1992, Neural Networks.

[16]  Razvan Pascanu,et al.  A simple neural network module for relational reasoning , 2017, NIPS.

[17]  Hugo Larochelle,et al.  Optimization as a Model for Few-Shot Learning , 2016, ICLR.

[18]  Samuel A. Nastase,et al.  Direct Fit to Nature: An Evolutionary Perspective on Biological and Artificial Neural Networks , 2019, Neuron.

[19]  Hassana K. Oyibo,et al.  Experience-dependent spatial expectations in mouse visual cortex , 2016, Nature Neuroscience.

[20]  Gašper Tkačik,et al.  Inferring the function performed by a recurrent neural network , 2019, PloS one.

[21]  Timothy P Lillicrap,et al.  Towards deep learning with segregated dendrites , 2016, eLife.

[22]  Pascal Poupart,et al.  Unsupervised Video Object Segmentation for Deep Reinforcement Learning , 2018, NeurIPS.

[23]  P. Best,et al.  Place cells and silent cells in the hippocampus of freely-behaving rats , 1989, The Journal of neuroscience : the official journal of the Society for Neuroscience.

[24]  Robert A. Legenstein,et al.  Long short-term memory and Learning-to-learn in networks of spiking neurons , 2018, NeurIPS.

[25]  Pieter R. Roelfsema,et al.  Control of synaptic plasticity in deep cortical networks , 2018, Nature Reviews Neuroscience.

[26]  Ronald J. Williams,et al.  Simple Statistical Gradient-Following Algorithms for Connectionist Reinforcement Learning , 2004, Machine Learning.

[27]  Geoffrey E. Hinton,et al.  The appeal of parallel distributed processing , 1986 .

[28]  Antonio Torralba,et al.  Comparison of deep neural networks to spatio-temporal cortical dynamics of human visual object recognition reveals hierarchical correspondence , 2016, Scientific Reports.

[29]  Ryota Tomioka,et al.  In Search of the Real Inductive Bias: On the Role of Implicit Regularization in Deep Learning , 2014, ICLR.

[30]  Jonas Kubilius,et al.  Brain-Score: Which Artificial Neural Network for Object Recognition is most Brain-Like? , 2018, bioRxiv.

[31]  A. Kitaoka,et al.  Illusory Motion Reproduced by Deep Neural Networks Trained for Prediction , 2018, Front. Psychol..

[32]  Winfried Denk,et al.  Progress and remaining challenges in high-throughput volume electron microscopy , 2018, Current Opinion in Neurobiology.

[33]  Timothy P Lillicrap,et al.  Deep Learning with Dynamic Spiking Neurons and Fixed Feedback Weights , 2017, Neural Computation.

[34]  Sergey Levine,et al.  Unsupervised Learning for Physical Interaction through Video Prediction , 2016, NIPS.

[35]  Kenneth D Harris,et al.  Challenges and opportunities for large-scale electrophysiology with Neuropixels probes , 2018, Current Opinion in Neurobiology.

[36]  David H. Wolpert,et al.  No free lunch theorems for optimization , 1997, IEEE Trans. Evol. Comput..

[37]  D. Field,et al.  Natural image statistics and efficient coding. , 1996, Network.

[38]  Timothy O'Leary,et al.  Fundamental bounds on learning performance in neural circuits , 2018, Proceedings of the National Academy of Sciences.

[39]  Alberto Testolin,et al.  Visual sense of number vs. sense of magnitude in humans and machines , 2020, Scientific Reports.

[40]  Timothy Lillicrap,et al.  Using Weight Mirrors to Improve Feedback Alignment , 2019 .

[41]  Ilya Sutskever,et al.  Language Models are Unsupervised Multitask Learners , 2019 .

[42]  Pieter R. Roelfsema,et al.  Attention-Gated Reinforcement Learning of Internal Representations for Classification , 2005, Neural Computation.

[43]  Rajesh P. N. Rao,et al.  Predictive Coding , 2019, A Blueprint for the Hard Problem of Consciousness.

[44]  Yoshua Bengio,et al.  Dendritic cortical microcircuits approximate the backpropagation algorithm , 2018, NeurIPS.

[45]  E. Marder,et al.  Central pattern generators and the control of rhythmic movements , 2001, Current Biology.

[46]  Anthony M. Zador.  A Critique of Pure Learning: What Artificial Neural Networks can Learn from Animal Brains , 2019 .

[47]  Konrad P. Kording,et al.  Neural spiking for causal inference , 2019 .

[48]  Kunihiko Fukushima,et al.  Neocognitron: A self-organizing neural network model for a mechanism of pattern recognition unaffected by shift in position , 1980, Biological Cybernetics.

[49]  Konrad P. Körding,et al.  What does it mean to understand a neural network? , 2019, ArXiv.

[50]  Konrad Paul Kording,et al.  Spiking allows neurons to estimate their causal effect , 2018, bioRxiv.

[51]  Qi Wang,et al.  Ligustilide improves aging-induced memory deficit by regulating mitochondrial related inflammation in SAMP8 mice , 2020, Aging.

[52]  M. A. MacIver,et al.  Neuroscience Needs Behavior: Correcting a Reductionist Bias , 2017, Neuron.

[53]  L. Abbott,et al.  Neural network dynamics. , 2005, Annual review of neuroscience.

[54]  栁下 祥.  A critical time window for dopamine actions on the structural plasticity of dendritic spines , 2016 .

[55]  Peter Elias,et al.  Predictive coding-I , 1955, IRE Trans. Inf. Theory.

[56]  Jim Williams,et al.  What Does It Mean? , 1907, California state journal of medicine.

[57]  Timothy P Lillicrap,et al.  Dendritic solutions to the credit assignment problem , 2019, Current Opinion in Neurobiology.

[58]  Kathleen E. Cullen,et al.  The vestibular system: multimodal integration and encoding of self-motion for motor control , 2012, Trends in Neurosciences.

[59]  Yoshua Bengio,et al.  Equilibrium Propagation: Bridging the Gap between Energy-Based Models and Backpropagation , 2016, Front. Comput. Neurosci..

[60]  Christoph Salge,et al.  Empowerment - an Introduction , 2013, ArXiv.

[61]  Konrad P. Körding,et al.  Toward an Integration of Deep Learning and Neuroscience , 2016, bioRxiv.

[62]  Geoffrey E. Hinton,et al.  Assessing the Scalability of Biologically-Motivated Deep Learning Algorithms and Architectures , 2018, NeurIPS.

[63]  Wulfram Gerstner,et al.  On the choice of metric in gradient-based theories of brain function , 2018, PLoS Comput. Biol..

[64]  Srinivas C. Turaga,et al.  Space-time wiring specificity supports direction selectivity in the retina , 2014, Nature.

[65]  Gašper Tkačik,et al.  Training and inferring neural network function with multi-agent reinforcement learning , 2020 .

[66]  Xiaohui Xie,et al.  Learning Curves for Stochastic Gradient Descent in Linear Feedforward Networks , 2003, Neural Computation.

[67]  Sophie Denève,et al.  Computational Account of Spontaneous Activity as a Signature of Predictive Coding , 2017, PLoS Comput. Biol..

[68]  Erich Elsen,et al.  Deep Speech: Scaling up end-to-end speech recognition , 2014, ArXiv.

[69]  Bartlett W. Mel,et al.  Pyramidal Neuron as Two-Layer Neural Network , 2003, Neuron.

[70]  Andrew M. Saxe,et al.  High-dimensional dynamics of generalization error in neural networks , 2017, Neural Networks.

[71]  Kevin M. Cury,et al.  DeepLabCut: markerless pose estimation of user-defined body parts with deep learning , 2018, Nature Neuroscience.

[72]  David J. Field,et al.  What Is the Other 85 Percent of V1 Doing , 2006 .

[73]  Marcel A. J. van Gerven,et al.  Deep Neural Networks Reveal a Gradient in the Complexity of Neural Representations across the Ventral Stream , 2014, The Journal of Neuroscience.

[74]  E. Oja.  Simplified neuron model as a principal component analyzer , 1982, Journal of mathematical biology.

[75]  Konrad P. Körding,et al.  Supervised and Unsupervised Learning with Two Sites of Synaptic Integration , 2001, Journal of Computational Neuroscience.

[76]  Yoshua Bengio,et al.  Scaling learning algorithms towards AI , 2007 .

[77]  J. DiCarlo,et al.  Using goal-driven deep learning models to understand sensory cortex , 2016, Nature Neuroscience.

[78]  Yoshua Bengio,et al.  Convolutional networks for images, speech, and time series , 1998 .

[79]  Byron M. Yu,et al.  Learning by neural reassociation , 2018, Nature Neuroscience.

[80]  Yann LeCun,et al.  Towards Understanding the Role of Over-Parametrization in Generalization of Neural Networks , 2018, ArXiv.

[81]  Chrystopher L. Nehaniv,et al.  Empowerment: a universal agent-centric measure of control , 2005, 2005 IEEE Congress on Evolutionary Computation.

[82]  Liam Paninski,et al.  Reinforcement Learning Recruits Somata and Apical Dendrites across Layers of Primary Sensory Cortex , 2019, Cell reports.

[83]  Allen Newell,et al.  GPS, a program that simulates human thought , 1995 .

[84]  Daniel L. K. Yamins,et al.  Deep Neural Networks Rival the Representation of Primate IT Cortex for Core Visual Object Recognition , 2014, PLoS Comput. Biol..

[85]  Nikolaus Kriegeskorte,et al.  Deep Supervised, but Not Unsupervised, Models May Explain IT Cortical Representation , 2014, PLoS Comput. Biol..

[86]  Loren L Looger,et al.  All‐optical functional synaptic connectivity mapping in acute brain slices using the calcium integrator CaMPARI , 2017, The Journal of physiology.

[87]  S. Scott.  Optimal feedback control and the neural basis of volitional motor control , 2004, Nature Reviews Neuroscience.

[88]  Pieter R. Roelfsema,et al.  A Biologically Plausible Learning Rule for Deep Learning in the Brain , 2018, ArXiv.

[89]  Andrew J King,et al.  Sensory cortex is optimized for prediction of future input , 2017, bioRxiv.

[90]  Surya Ganguli,et al.  Task-Driven Convolutional Recurrent Models of the Visual System , 2018, NeurIPS.

[91]  Yoshua Bengio,et al.  Attention-Based Models for Speech Recognition , 2015, NIPS.

[92]  Jason Yosinski,et al.  Understanding Neural Networks via Feature Visualization: A survey , 2019, Explainable AI.

[93]  Daniel L. K. Yamins,et al.  A Task-Optimized Neural Network Replicates Human Auditory Behavior, Predicts Brain Responses, and Reveals a Cortical Processing Hierarchy , 2018, Neuron.

[94]  Katie C. Bittner,et al.  Behavioral time scale synaptic plasticity underlies CA1 place fields , 2017, Science.

[95]  James J DiCarlo,et al.  Neural population control via deep image synthesis , 2018, Science.

[96]  M. Botvinick,et al.  The successor representation in human reinforcement learning , 2016, Nature Human Behaviour.

[97]  Yali Amit,et al.  Deep Learning With Asymmetric Connections and Hebbian Updates , 2018, Front. Comput. Neurosci..

[98]  Rafal Bogacz,et al.  An Approximation of the Error Backpropagation Algorithm in a Predictive Coding Network with Local Hebbian Synaptic Plasticity , 2017, Neural Computation.

[99]  David C. Van Essen,et al.  Information Processing Strategies and Pathways in the Primate Visual System . , 1995 .

[100]  Anthony M. Zador,et al.  A critique of pure learning and what artificial neural networks can learn from animal brains , 2019, Nature Communications.

[101]  Colin J. Akerman,et al.  Random synaptic feedback weights support error backpropagation for deep learning , 2016, Nature Communications.

[102]  Richard S. Zemel,et al.  Prototypical Networks for Few-shot Learning , 2017, NIPS.

[103]  Yang Gao,et al.  Deep learning for tactile understanding from visual and haptic data , 2015, 2016 IEEE International Conference on Robotics and Automation (ICRA).

[104]  Leena E Williams,et al.  Higher-Order Thalamocortical Inputs Gate Synaptic Long-Term Potentiation via Disinhibition , 2018, Neuron.

[105]  T. Vogels,et al.  Synaptic Transmission Optimization Predicts Expression Loci of Long-Term Plasticity , 2017, Neuron.

[106]  Jane X. Wang,et al.  Reinforcement Learning, Fast and Slow , 2019, Trends in Cognitive Sciences.

[107]  Geoffrey E. Hinton,et al.  ImageNet classification with deep convolutional neural networks , 2012, Commun. ACM.

[108]  David J. Freedman,et al.  Inferring learning rules from distribution of firing rates in cortical neurons , 2015, Nature Neuroscience.

[109]  J. Kwag,et al.  The timing of external input controls the sign of plasticity at local synapses , 2009, Nature Neuroscience.