Biological_Learning code for CIFAR

hi, I think your work is very interesting. I'd like to try it on more complex databases. So would you also share the code for CIFAR? It's a good starting point for me. Thanks a lot in advance.

Mar 07 '21 07:03 lqiang2003cn

Hi. In our fork, we have implemented the CIFAR version--as well as two other tasks. Hope this helps!

Apr 16 '22 07:04 heydarshahi

hi, i don't quite understand the above equation. Would anyone explain it in detail for me? i am trying to explore the online algorithm which i think might be useful in the filed of continual learning. Any suggestions are welcome.

Aug 11 '22 08:08 lqiang2003cn

Hi lqiang2003cn this equation makes it possible to compute the activations of the hidden units in the network. The hidden units here mutually inhibit each other through the term proportional to $w_\text{inh}$. The term $I_\mu$ is the input to this network motif, and the last term is the standard decay term in firing rate models. This equation is a "biological" way of implementing the algorithm that we've discussed in the paper. For practical use, it is better (because it works faster) to use the method described in the section "A Fast Implementation", see equation (10).

Aug 15 '22 16:08 DimaKrotov