Deep Learning without Weight Transport
This work addresses a key problem in neuroscience and AI by enabling biologically plausible deep learning, though it is incremental as it builds on existing feedback alignment methods.
The paper tackles the biological implausibility of weight transport in deep learning by proposing two mechanisms—a weight mirror circuit and a modified Kolen-Pollack algorithm—that enable feedback paths to learn appropriate synaptic weights without weight transport. On ImageNet, these mechanisms outperform feedback alignment and sign-symmetry methods and nearly match backpropagation, achieving results close to the standard algorithm.
Current algorithms for deep learning probably cannot run in the brain because they rely on weight transport, where forward-path neurons transmit their synaptic weights to a feedback path, in a way that is likely impossible biologically. An algorithm called feedback alignment achieves deep learning without weight transport by using random feedback weights, but it performs poorly on hard visual-recognition tasks. Here we describe two mechanisms - a neural circuit called a weight mirror and a modification of an algorithm proposed by Kolen and Pollack in 1994 - both of which let the feedback path learn appropriate synaptic weights quickly and accurately even in large networks, without weight transport or complex wiring.Tested on the ImageNet visual-recognition task, these mechanisms outperform both feedback alignment and the newer sign-symmetry method, and nearly match backprop, the standard algorithm of deep learning, which uses weight transport.