Topic: Deep Feedback Control

A key driver of mammalian intelligence is the ability to represent sensory inputs at different levels of abstraction. In deep learning, hierarchical neural networks learn such abstractions through top-down signals that drive synaptic plasticity. This is commonly done with backpropagation, an algorithm that is considered biologically implausible. In this work, we present a novel dynamic target propagation framework for hierarchical learning that uses Differential Hebbian (DH) updates, a rate-based version of Spike-Time-Dependent Plasticity. Our framework is based on the Deep Feedback Control (D

The success of deep learning sparked interest in whether the brain similarly learns its hierarchical representations. However, current biologically-plausible models for hierarchical credit assignment (HCA) --i.e., determining how to adjust synapses across hierarchies-- assume that the effect of feedback on forward processing is negligible. This weak feedback assumption is problematic in biologically realistic noisy environments and is at odds with experimental evidence showing that the effect of feedback on neural activities can be strong. To overcome this limitation, we revisit the recent Dee