This repository is meant to simulate small clusters of neurons preforming specific tasks. These clusters, called circuit objects, are comprised of node objects that take one or more linear inputs and send multiple outputs to other nodes in a circuit. Circuits are trained in groups of generations by a trainer object using a genetic algorithm to learn from a data set, and create a better preforming next generation. Circuits can be individually written or read to .json files using Google's gson serialization & deserialization library. The end goal for this repository is to create both a method to train circuits to preform tasks efficiently and be able to visualize the learning process and the result of it.
This version will be the MVP (minimum viable product). It will have the baseline features to train circuits according to a set of inputs, visualize a circuit, read/write circuits to files, and utilize multitasking. It will be packaged into a .jar file to be used on any JVM, with dedicated support for windows and prespective support on Linux command line for a Raspbian Lite x64 server.
This version will build on the MVP to deliver easier implementation of real-time or feedback reliant training for circuits using a physics or other type of engine.
This version will add a better UI to make the visualization easier on windows.