What would it look like to add the backprop step as Python too? #22
Labels
enhancement
New feature or request
help wanted
Extra attention is needed
question
Further information is requested
We have the forward pass, which already makes the nice point that the network is a function and the forward pass is cheap... but this does not reflect hyperparams directly, eg how regularization works.
What would it look like to add the training loop as a function as well? Too much?
Moved from old project.
The text was updated successfully, but these errors were encountered: