Ari Sanguinetti
Ari Sanguinetti
I'm a noobie on testing, not even sure if what I made is correct. I've created a random NN and saved the initial status. then trained the NN 100 times...
I think this will improve the quality of this repo, maybe @shiffman can do a video about lintering ;)
Add the option to decay the learning rate based on number of trainings
I've added some visualization on the NN status and predictions. I was inspired by [this article](https://ml4a.github.io/ml4a/looking_inside_neural_nets/) on ML4A. Also, I've reduced strokeWeight since it will provide a better result (when...