Good catch. Do you mind submitting a PR with the working version?

from nn-from-scratch.

I tried to do it but doing the modification in Jupiter creates a lot of other changes in the file beyond adding the three lines that follow. As I'm using Python v3.x, I don't wan't to generate more issues...:

Gradient descent parameters (I picked these by hand)

epsilon = 0.01 # learning rate for gradient descent
reg_lambda = 0.01 # regularization strength

from nn-from-scratch.

All the other changes are probably related to the cell metadata used for the notebook (collapsed, execution_count...).
Anyway, #5 should fix it! (and the diff is kept to the bare minimum).

from nn-from-scratch.

Thanks! Yeah, the other changes are the probably the cached results in the notebook. Could also just clean the output after changing the code. Anyway, I'm closing this. Thanks for the PR.

from nn-from-scratch.