NeuralNetworkGA1

Neural Network with Genetic Algorithms Experiment 1 (food and water) with online demo!

An experiment using Neural Networks and genetic algorithms I have always found you understand something better if you build it yourself.

See it running at https://youtu.be/yZUhprqGVTE

Live demo at http://experiments.robsmithdev.co.uk/NeuralNetworkGA1/

My aim was to make a "lifeform" that learnt basic needs. The lifeform needs water and food to survive. Water is always available, food is scattered and once eaten re-spawns elsewhere. If the "lifeform" runs out of food or water then it is classed as "dead"

There are 30 "lifeforms" in each generation, and each generation runs for a maximum of 6000 iterations. The project is setup with a Neural Network with 8 inputs, two hidden layers and 2 outputs.

With the current configuration and topology, by the 950th generation 2/3 of the lifeforms survive.

I have included data for the network after I left it running over night (it reached 17120 generations)

I plan to use this project as a starting point for experiment 2, which will be more complex

I found the following resources useful and inspirational while creating this:

A colleague: https://github.com/GuillaumeBouchetEpitech/

MIT AI Courseware - Worth a watch: https://www.youtube.com/playlist?list=PLUl4u3cNGP63gFHB6xb-kVBiQHYe_4hSi

Good resource and tutorials: http://www.ai-junkie.com/