BaZi approximator is a playground project for experimenting with simple neural networks.
It uses Synaptic JS library.
In order to facilitate experimentation and training on a large dataset without loosing data, it saves the network setup every few thousand iterations and reloads on every start.