English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

The backpropagation algorithm has experienced remarkable success in traininglarge-scale artificial neural networks; however, its biological plausibilityhas been strongly criticized, and it remains an open question whether the brainemploys supervised learning mechanisms akin to it. Here, we propose correlativeinformation maximization between layer activations as an alternative normativeapproach to describe the signal propagation in biological neural networks inboth forward and backward directions. This new framework addresses manyconcerns about the biological-plausibility of conventional artificial neuralnetworks and the backpropagation algorithm. The coordinate descent-basedoptimization of the corresponding objective, combined with the mean squareerror loss function for fitting labeled supervision data, gives rise to aneural network structure that emulates a more biologically realistic network ofmulti-compartment pyramidal neurons with dendritic processing and lateralinhibitory neurons. Furthermore, our approach provides a natural resolution tothe weight symmetry problem between forward and backward signal propagationpaths, a significant critique against the plausibility of the conventionalbackpropagation algorithm. This is achieved by leveraging two alternative, yetequivalent forms of the correlative mutual information objective. Thesealternatives intrinsically lead to forward and backward prediction networkswithout weight symmetry issues, providing a compelling solution to thislong-standing challenge.

Correlative Information Maximization: A Biologically Plausible Approach to Supervised Deep Neural Networks without Weight Symmetry