Python Neural Network module

Experimental module without dependencies, using only python standard libraries

• Written in Python 3.x
• User configurable network architecture
• User configurable network training parameters
• No external libraries (like numpy)

I have written this module to study the algorithms of a multilayer perceptron. The goal is to keep the code as simple as possible to clearly see network algorithms without focusing too much on performance.

To do:

• Normalize the input of the network
• Make a loader to feed the network from .csv/.xml/... files
• Stop training by timeout or error stabilization
• Make a Jupyter notebook with the training of a network with real world public datasets

How to run the demo:

python RNA.py