Dendritic cortical microcircuits approximate the backpropagation algorithm

Deep learning has seen remarkable developments over the last years, many of them inspired by neuroscience. However, the main learning mechanism behind these advances - error backpropagation - appears to be at odds with neurobiology. Here, we introduce a multilayer neuronal network model with simplified dendritic compartments in which error-driven synaptic plasticity adapts the network towards a global desired output. In contrast to previous work our model does not require separate phases and synaptic learning is driven by local dendritic prediction errors continuously in time. Such errors originate at apical dendrites and occur due to a mismatch between predictive input from lateral interneurons and activity from actual top-down feedback. Through the use of simple dendritic compartments and different cell-types our model can represent both error and normal activity within a pyramidal neuron. We demonstrate the learning capabilities of the model in regression and classification tasks, and show analytically that it approximates the error backpropagation algorithm. Moreover, our framework is consistent with recent observations of learning between brain areas and the architecture of cortical microcircuits. Overall, we introduce a novel view of learning on dendritic cortical circuits and on how the brain may solve the long-standing synaptic credit assignment problem.

[1]  Georg B. Keller,et al.  Mismatch Receptive Fields in Mouse Visual Cortex , 2016, Neuron.

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

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

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

[5]  Claudia Clopath,et al.  Modeling somatic and dendritic spike mediated plasticity at the single neuron and network level , 2017, Nature Communications.

[6]  Konrad P. Kording,et al.  Towards an integration of deep learning and neuroscience , 2016, bioRxiv.

[7]  S. Manita,et al.  A Top-Down Cortical Circuit for Accurate Sensory Perception , 2015, Neuron.

[8]  W. Gerstner,et al.  Connectivity reflects coding: a model of voltage-based STDP with homeostasis , 2010, Nature Neuroscience.

[9]  Stephen Grossberg,et al.  Competitive Learning: From Interactive Activation to Adaptive Resonance , 1987, Cogn. Sci..

[10]  Georg B. Keller,et al.  Learning Enhances Sensory and Multiple Non-sensory Representations in Primary Visual Cortex , 2015, Neuron.

[11]  Francis Crick,et al.  The recent excitement about neural networks , 1989, Nature.

[12]  Caspar M. Schwiedrzik,et al.  High-Level Prediction Signals in a Low-Level Area of the Macaque Face-Processing Hierarchy , 2017, Neuron.

[13]  M. Larkum A cellular mechanism for cortical associations: an organizing principle for the cerebral cortex , 2013, Trends in Neurosciences.

[14]  T. Lillicrap,et al.  Preference Distributions of Primary Motor Cortex Neurons Reflect Control Solutions Optimized for Limb Biomechanics , 2013, Neuron.

[15]  Yoshua Bengio,et al.  Difference Target Propagation , 2014, ECML/PKDD.

[16]  Geoffrey E. Hinton,et al.  Learning Representations by Recirculation , 1987, NIPS.

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

[18]  Geoffrey E. Hinton,et al.  A Learning Algorithm for Boltzmann Machines , 1985, Cogn. Sci..

[19]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[20]  Ha Hong,et al.  Performance-optimized hierarchical models predict neural responses in higher visual cortex , 2014, Proceedings of the National Academy of Sciences.

[21]  R. Froemke Plasticity of cortical excitatory-inhibitory balance. , 2015, Annual review of neuroscience.

[22]  K. Svoboda,et al.  The subcellular organization of neocortical excitatory connections , 2009, Nature.

[23]  Walter Senn,et al.  Spatio-Temporal Credit Assignment in Neuronal Population Learning , 2011, PLoS Comput. Biol..

[24]  Michael P Stryker,et al.  A cortical disinhibitory circuit for enhancing adult plasticity , 2015, eLife.

[25]  Lin Tian,et al.  Activity in motor-sensory projections reveals distributed coding in somatosensation , 2012, Nature.

[26]  Pierre Priouret,et al.  Adaptive Algorithms and Stochastic Approximations , 1990, Applications of Mathematics.

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

[28]  Takaki Komiyama,et al.  Learning enhances the relative impact of top-down processing in the visual cortex , 2015, Nature Neuroscience.

[29]  Michael J. Higley,et al.  Input-Specific NMDAR-Dependent Potentiation of Dendritic GABAergic Inhibition , 2017, Neuron.

[30]  Xiaohui Xie,et al.  Equivalence of Backpropagation and Contrastive Hebbian Learning in a Layered Network , 2003, Neural Computation.

[31]  Alexander Attinger,et al.  Visuomotor Coupling Shapes the Functional Development of Mouse Visual Cortex , 2017, Cell.

[32]  Konrad Paul Kording,et al.  Learning with two sites of synaptic integration , 2000, Network.

[33]  P. J. Sjöström,et al.  Rate, Timing, and Cooperativity Jointly Determine Cortical Synaptic Plasticity , 2001, Neuron.

[34]  Arild Nøkland,et al.  Direct Feedback Alignment Provides Learning in Deep Neural Networks , 2016, NIPS.

[35]  Karl J. Friston,et al.  A theory of cortical responses , 2005, Philosophical Transactions of the Royal Society B: Biological Sciences.

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

[37]  N. Spruston Pyramidal neurons: dendritic structure and synaptic integration , 2008, Nature Reviews Neuroscience.

[38]  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.

[39]  Nando de Freitas,et al.  Cortical microcircuits as gated-recurrent neural networks , 2017, NIPS.

[40]  James L. McClelland,et al.  What Learning Systems do Intelligent Agents Need? Complementary Learning Systems Theory Updated , 2016, Trends in Cognitive Sciences.

[41]  Yann Le Cun,et al.  A Theoretical Framework for Back-Propagation , 1988 .

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

[43]  Randall C. O'Reilly,et al.  Biologically Plausible Error-Driven Learning Using Local Activation Differences: The Generalized Recirculation Algorithm , 1996, Neural Computation.

[44]  Henning Sprekeler,et al.  Inhibitory Plasticity Balances Excitation and Inhibition in Sensory Pathways and Memory Networks , 2011, Science.

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

[46]  Christoph E. Schreiner,et al.  Developmental sensory experience balances cortical excitation and inhibition , 2010, Nature.

[47]  Yotam Luz,et al.  Balancing Feed-Forward Excitation and Inhibition via Hebbian Inhibitory Synaptic Plasticity , 2012, PLoS Comput. Biol..

[48]  Georg B. Keller,et al.  A Sensorimotor Circuit in Mouse Cortex for Visual Flow Predictions , 2017, Neuron.

[49]  Alison L. Barth,et al.  Somatostatin-expressing neurons in cortical networks , 2016, Nature Reviews Neuroscience.

[50]  Yoshua Bengio,et al.  How Auto-Encoders Could Provide Credit Assignment in Deep Networks via Target Propagation , 2014, ArXiv.

[51]  W. Senn,et al.  Learning by the Dendritic Prediction of Somatic Spiking , 2014, Neuron.

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

[53]  Timothée Masquelier,et al.  Unsupervised Learning of Visual Features through Spike Timing Dependent Plasticity , 2007, PLoS Comput. Biol..

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

[55]  Sander W. Keemink,et al.  Behavioral-state modulation of inhibition is context-dependent and cell type specific in mouse visual cortex , 2016, eLife.

[56]  Geoffrey E. Hinton,et al.  Learning representations by back-propagating errors , 1986, Nature.

[57]  Rajesh P. N. Rao,et al.  Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects. , 1999 .

[58]  James L. McClelland,et al.  Why there are complementary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory. , 1995, Psychological review.

[59]  Wolfgang Maass,et al.  Bayesian Computation Emerges in Generic Cortical Microcircuits through Spike-Timing-Dependent Plasticity , 2013, PLoS Comput. Biol..