Towards Biologically Plausible Deep Learning

Abstract

Neuroscientists have long criticised deep learning algorithms as incompatible with current knowledge of neurobiology. We explore more biologically plausible versions of deep representation learning, focusing here mostly on unsupervised learning but developing a learning mechanism that could account for supervised, unsupervised and reinforcement learning. The starting point is that the basic learning rule believed to govern synaptic weight updates (Spike-Timing-Dependent Plasticity) arises out of a simple update rule that makes a lot of sense from a machine learning point of view and can be interpreted as gradient descent on some objective function so long as the neuronal dynamics push firing rates towards better values of the objective function (be it supervised, unsupervised, or reward-driven). The second main idea is that this corresponds to a form of the variational EM algorithm, i.e., with approximate rather than exact posteriors, implemented by neural dynamics. Another contribution of this paper is that the gradients required for updating the hidden states in the above variational interpretation can be estimated using an approximation that only requires propagating activations forward and backward, with pairs of layers learning to form a denoising auto-encoder. Finally, we extend the theory about the probabilistic interpretation of auto-encoders to justify improved sampling schemes based on the generative interpretation of denoising auto-encoders, and we validate all these ideas on generative learning tasks.

References in corpus (5)

• Semi-Supervised Learning with Deep Generative Models
• Markov Chain Monte Carlo and Variational Inference: Bridging the Gap
• Fast Inference in Sparse Coding Algorithms with Applications to Object Recognition
• How Auto-Encoders Could Provide Credit Assignment in Deep Networks via Target Propagation
• Random feedback weights support learning in deep neural networks

Cited by in corpus (62)

• Deep Learning in Spiking Neural Networks
• An Introduction to Deep Reinforcement Learning
• STDP-based spiking deep convolutional neural networks for object recognition
• Direct Feedback Alignment Provides Learning in Deep Neural Networks
• Bottom-up and top-down approaches for the design of neuromorphic processing systems: Tradeoffs and synergies between natural and artificial intelligence
• Learning Longer-term Dependencies in RNNs with Auxiliary Losses
• Demonstration of Decentralized, Physics-Driven Learning
• Putting An End to End-to-End: Gradient-Isolated Learning of Representations
• Towards an integration of deep learning and neuroscience
• How Important is Weight Symmetry in Backpropagation?
• Sparse Unsupervised Capsules Generalize Better
• A Theoretical Framework for Target Propagation
• Understanding Synthetic Gradients and Decoupled Neural Interfaces
• SoftHebb: Bayesian Inference in Unsupervised Hebbian Soft Winner-Take-All Networks
• A bio-inspired bistable recurrent cell allows for long-lasting memory
• Revisiting Locally Supervised Learning: an Alternative to End-to-end Training
• Learning with hidden variables
• Deep Learning-Assisted Classification of Site-Resolved Quantum Gas Microscope Images
• Morphological Error Detection in 3D Segmentations
• LoCo: Local Contrastive Representation Learning
• ETLP: Event-based Three-factor Local Plasticity for online learning with neuromorphic hardware
• Training recurrent networks to generate hypotheses about how the brain solves hard navigation problems
• Early Inference in Energy-Based Models Approximates Back-Propagation
• Invariant recognition drives neural representations of action sequences
• Equilibrium Propagation: Bridging the Gap Between Energy-Based Models and Backpropagation
• SpikePropamine: Differentiable Plasticity in Spiking Neural Networks
• Learning to Adapt by Minimizing Discrepancy
• Sparsely Activated Networks
• Understanding Important Features of Deep Learning Models for Transmission Electron Microscopy Image Segmentation
• Transfer entropy-based feedback improves performance in artificial neural networks
• Efficient Training of Very Deep Neural Networks for Supervised Hashing
• CSNNs: Unsupervised, Backpropagation-free Convolutional Neural Networks for Representation Learning
• Adaptive Bidirectional Backpropagation: Towards Biologically Plausible Error Signal Transmission in Neural Networks
• Object Based Attention Through Internal Gating
• A Neural Spiking Approach Compared to Deep Feedforward Networks on Stepwise Pixel Erasement
• DANTE: Deep AlterNations for Training nEural networks
• Optimality of short-term synaptic plasticity in modelling certain dynamic environments
• Giving Up Control: Neurons as Reinforcement Learning Agents
• Backprop-Q: Generalized Backpropagation for Stochastic Computation Graphs
• Closing the loop on multisensory interactions: A neural architecture for multisensory causal inference and recalibration
• Data augmentation and image understanding
• Backpropagation and Biological Plausibility
• Ghost Units Yield Biologically Plausible Backprop in Deep Neural Networks
• Meta-Learning Bidirectional Update Rules
• Local learning rules to attenuate forgetting in neural networks
• MAP Propagation Algorithm: Faster Learning with a Team of Reinforcement Learning Agents
• Questions to Guide the Future of Artificial Intelligence Research
• Deep learning for image segmentation: veritable or overhyped?
• Tourbillon: a Physically Plausible Neural Architecture
• Invertible generalized synchronization: A putative mechanism for implicit learning in biological and artificial neural systems
• Learning in Deep Neural Networks Using a Biologically Inspired Optimizer
• Towards Robust Pattern Recognition: A Review
• Understanding Information Processing in Human Brain by Interpreting Machine Learning Models
• Associated Learning: Decomposing End-to-end Backpropagation based on Auto-encoders and Target Propagation
• A Threshold-based Scheme for Reinforcement Learning in Neural Networks
• Sparsely Activated Networks: A new method for decomposing and compressing data
• Biologically Plausible Sequence Learning with Spiking Neural Networks
• Learning spatial hearing via innate mechanisms
• TEDL: A Text Encryption Method Based on Deep Learning
• Spatiotemporal Local Propagation
• Dynamical Mechanism of Sampling-based Stochastic Inference under Probabilistic Population Codes
• Perspective: Purposeful Failure in Artificial Life and Artificial Intelligence