How Auto-Encoders Could Provide Credit Assignment in Deep Networks via Target Propagation

Abstract

We propose to exploit {\em reconstruction} as a layer-local training signal for deep learning. Reconstructions can be propagated in a form of target propagation playing a role similar to back-propagation but helping to reduce the reliance on derivatives in order to perform credit assignment across many levels of possibly strong non-linearities (which is difficult for back-propagation). A regularized auto-encoder tends produce a reconstruction that is a more likely version of its input, i.e., a small move in the direction of higher likelihood. By generalizing gradients, target propagation may also allow to train deep networks with discrete hidden units. If the auto-encoder takes both a representation of input and target (or of any side information) in input, then its reconstruction of input representation provides a target towards a representation that is more likely, conditioned on all the side information. A deep auto-encoder decoding path generalizes gradient propagation in a learned way that can could thus handle not just infinitesimal changes but larger, discrete changes, hopefully allowing credit assignment through a long chain of non-linear operations. In addition to each layer being a good auto-encoder, the encoder also learns to please the upper layers by transforming the data into a space where it is easier to model by them, flattening manifolds and disentangling factors. The motivations and theoretical justifications for this approach are laid down in this paper, along with conjectures that will have to be verified either mathematically or experimentally, including a hypothesis stating that such auto-encoder mediated target propagation could play in brains the role of credit assignment through many non-linear, noisy and discrete transformations.

References in corpus (11)

• Auto-Encoding Variational Bayes
• Stochastic Backpropagation and Approximate Inference in Deep Generative Models
• Better Mixing via Deep Representations
• Knowledge Matters: Importance of Prior Information for Optimization
• Deep Learning of Representations: Looking Forward
• Deep AutoRegressive Networks
• A Deep and Tractable Density Estimator
• Efficient Gradient-Based Inference through Transformations between Bayes Nets and Neural Nets
• Reweighted Wake-Sleep
• From Machine Learning to Machine Reasoning
• Multimodal Transitions for Generative Stochastic Networks

Cited by in corpus (42)

• NICE: Non-linear Independent Components Estimation
• Semi-Supervised Learning with Ladder Networks
• Deep Predictive Coding Networks for Video Prediction and Unsupervised Learning
• Towards Biologically Plausible Deep Learning
• Towards an integration of deep learning and neuroscience
• Decoupled Neural Interfaces using Synthetic Gradients
• How Important is Weight Symmetry in Backpropagation?
• Deconstructing the Ladder Network Architecture
• Feedback alignment in deep convolutional networks
• Dendritic error backpropagation in deep cortical microcircuits
• A Theoretical Framework for Target Propagation
• Principled Training of Neural Networks with Direct Feedback Alignment
• Convergent Block Coordinate Descent for Training Tikhonov Regularized Deep Neural Networks
• Two Routes to Scalable Credit Assignment without Weight Symmetry
• Conducting Credit Assignment by Aligning Local Representations
• LoCo: Local Contrastive Representation Learning
• Deep Directed Generative Autoencoders
• Deriving Differential Target Propagation from Iterating Approximate Inverses
• Kernelized information bottleneck leads to biologically plausible 3-factor Hebbian learning in deep networks
• Deep Learning as a Mixed Convex-Combinatorial Optimization Problem
• On Training Implicit Models
• GAIT-prop: A biologically plausible learning rule derived from backpropagation of error
• Cortical credit assignment by Hebbian, neuromodulatory and inhibitory plasticity
• Gradient-Free Training of Recurrent Neural Networks using Random Perturbations
• Training Language Models Using Target-Propagation
• Biological credit assignment through dynamic inversion of feedforward networks
• Learning to Learn with Feedback and Local Plasticity
• Identifying Learning Rules From Neural Network Observables
• Gradient target propagation
• DANTE: Deep AlterNations for Training nEural networks
• Large-Scale Gradient-Free Deep Learning with Recursive Local Representation Alignment
• Direct Feedback Alignment Scales to Modern Deep Learning Tasks and Architectures
• Credit Assignment Through Broadcasting a Global Error Vector
• On Kernel Method-Based Connectionist Models and Supervised Deep Learning Without Backpropagation
• Meta-Learning Bidirectional Update Rules
• Benchmarking the Accuracy and Robustness of Feedback Alignment Algorithms
• Biologically-Motivated Deep Learning Method using Hierarchical Competitive Learning
• How and When Random Feedback Works: A Case Study of Low-Rank Matrix Factorization
• Learning to infer in recurrent biological networks
• Associated Learning: Decomposing End-to-end Backpropagation based on Auto-encoders and Target Propagation
• Self Normalizing Flows
• Target Propagation via Regularized Inversion