Package arranged to test 1001 different neural nets on the MNIST dataset.
I used to have a Heroku dyno running visualizations of the network, but that
expired :(, but for now you have rely on the summary blog posts:
https://ahgamut.github.io/2020/11/20/netpicking-1/ https://ahgamut.github.io/2020/11/24/netpicking-2/ https://ahgamut.github.io/2021/11/24/netpicking-3/
What if I had a bunch of different neural nets all trained on the same data?
The package contains 1001 different neural nets written in Pytorch, along with
scaffolding to test them on the MNIST dataset. The mnistk-webapp
repository contains a web application deployable to Heroku, to see how each
network performed.