A basic neural network implementing the concepts of select - cross - mutate to generate the next generation
This project has been made compatible with the java api provided by Dr Tim French.
The model is created using tensor flow and trains by interacting with the api via a java server that must be run before execution.
A future release will contain an agent capable of running the api just off saved model files.
pip3 install -r requirements.txt to install
run installer.py with python3 installer.py to install dependencies and compile java
WARNING script will fail if jdk or svn are not in path
download and extract the jni for tensorflow appropriate to your system and place it in the Tensorflow-Config directory
python3 start_server.py to start the server
then in a seperate shell python3 train_agent.py to train the model
Currently hyperparameters are all manually set inside train_agent.py
Support for custom hyperparameters and appropriate model identification is planned