Search:
Match:
1 results

Analysis

This paper addresses the biological implausibility of Backpropagation Through Time (BPTT) in training recurrent neural networks. It extends the E-prop algorithm, which offers a more biologically plausible alternative to BPTT, to handle deep networks. This is significant because it allows for online learning of deep recurrent networks, mimicking the hierarchical and temporal dynamics of the brain, without the need for backward passes.
Reference

The paper derives a novel recursion relationship across depth which extends the eligibility traces of E-prop to deeper layers.