Skip to content
Backpropogation with random weights
Python
Branch: master
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
.idea
notes.md
random_feedback_weights.py
readme.md

readme.md

Recently, I came across a paper called Random feedback weights support learning in deep neural networks. I found it very fascinating that they were able to match the performance of backpropagation algorithm by propagating the error through random weights! (instead of the same weights they used for forward prop) Yes random weights! The idea is inspired by the biological limitation of the brain to deliver exact backpropagation information.

I decided to replicate this paper and take a closer look for myself. Here is what I found out: http://www.siarez.com/projects/random-backpropogation

You can’t perform that action at this time.