GeneticAlgorithm

Working my way to creating an version of Ragnarok Online that runs a genetic algorithm. lol. Check out the journey at: https://dataontherocks.wordpress.com/2017/06/29/simulating-multi-agent-survival-using-neuroevolutiongenetic-algorithms-python-part-1/

1. shakespeare_monkey_thecodingtrain.py

- inspired by: The Coding Train's https://www.youtube.com/watch?v=-jv3CgDN9sc&list=PLRqwX-V7Uu6bJM3VgzjNV5YxVxUwzALHV&index=4

2. agent_survival.py

- inspired by:
    - The Coding Train's https://www.youtube.com/watch?v=-jv3CgDN9sc&list=PLRqwX-V7Uu6bJM3VgzjNV5YxVxUwzALHV&index=4
    - emgoz's https://github.com/emgoz/Neural-network-snake
    - Bart van Dooren's https://github.com/Nerfling/Evolving-Creatures-Using-Neural-Networks-And-Genetic-Algorithms
    - stela zhang's https://www.youtube.com/watch?v=eZ14la6zttM
    - phyces' https://www.youtube.com/watch?v=GvEywP8t12I
    - Geoff. S. Nitschke and Leo. H. Langenhoven's Neuro-Evolution for Competitive Co-evolution of Biologically Canonical Predator and Prey Behaviors (https://people.cs.uct.ac.za/~gnitschke/projects/papers/2010-Neuro-Evolution%20for%20Competitive%20Co-evolution%20of%20Biologically%20Canonical%20Predator%20and%20Prey%20Behaviors.pdf)
    
- implemented using pybrain and pygame
    - pybrain: http://pybrain.org/docs/
    - pygame: http://thepythongamebook.com
        - Most of the implementations I found were on Java or C. That being said, it was quite a challenge to make pygame work. lol.

- Notes (or questions I was able to answer through study):
    - Important to understand the difference of a normal feedforward neural net (with backprop) and a feedforward neural net that is optimized using genetic algorithms/neuroevolution. 
    - What does it mean to evolve a neural net based on optimal behavior?
    - Is genome always a list when being subjected to crossovers? or can different parts of the genome have its own individual crossover instance?
    - Are there different ways of doing natural selection, generation, crossover, and mutation?
    - How should the fitness function be created to steer agents towards a certain direction of optimal behavior?
    - Steering is currently brute forced... will have to be revised.