Been working on some improved Neural Network visualisation in Clojure.
Here’s a view of a trained 4-layer classifier. The bright yellow dot at the top indicates that the network has classified this particular training case into the 4th class.
The visualisation is actually live: you can see the weights and activations change as you run through different cases and train the network.
The code is part of the “Nurokit” framework that I’m slowly pushing towards a usable state:
Looking forward to doing a lot more machine learning visualisations in the next few weeks!