If it's not too much, I have a suggestion for a cleaner implementation of x_vals normalization.
Instead of defining a separate function, and since the objective of normalization is just to allow for faster convergence, we could perform the following:
nfmcclure
commented
6 years ago
Hi,
If it's not too much, I have a suggestion for a cleaner implementation of x_vals normalization.
Instead of defining a separate function, and since the objective of normalization is just to allow for faster convergence, we could perform the following:
x_vals_train = x_vals_train / x_vals_train.max(axis=0) x_vals_test = x_vals_test / x_vals_test.max(axis=0)
This implementation is cleaner and achieve similar loss results. Do let me know what you think.
Sincerely, Yi Xiang chongyixiang@gmail.com