Local plasticity rules can learn deep representations using self-supervised contrastive predictions
This addresses the challenge of understanding brain-like learning for researchers in computational neuroscience and AI, offering a biologically constrained method for deep learning, though it appears incremental as it builds on existing contrastive and local learning ideas.
The authors tackled the problem of learning deep hierarchical representations in biologically plausible neural networks by proposing a local, Hebbian learning rule inspired by neuroscience and self-supervised contrastive predictions, which builds deep representations for images, speech, and video without backpropagation.
Learning in the brain is poorly understood and learning rules that respect biological constraints, yet yield deep hierarchical representations, are still unknown. Here, we propose a learning rule that takes inspiration from neuroscience and recent advances in self-supervised deep learning. Learning minimizes a simple layer-specific loss function and does not need to back-propagate error signals within or between layers. Instead, weight updates follow a local, Hebbian, learning rule that only depends on pre- and post-synaptic neuronal activity, predictive dendritic input and widely broadcasted modulation factors which are identical for large groups of neurons. The learning rule applies contrastive predictive learning to a causal, biological setting using saccades (i.e. rapid shifts in gaze direction). We find that networks trained with this self-supervised and local rule build deep hierarchical representations of images, speech and video.