Experiments

These are the list of experiments I want to work on, in order to develop my ideas. I will divide them into phases, to provide a gradual buildup of capacity and ideas.

The majority of these experiments are based on evolutionary algorithms and games / setups to test different ideas and hypotheses.

Phase 1: basics

Objectives

1. Implement and comprehend the key EA optimization algorithms

2. Build an experimental setup: algorithms / environments / monitoring / analysis

3. Improve my capability on designing environments / games to test ideas

4. Focus mainly on simulation and grid worlds

5. Get a process in place: experimentation, writing and video editing

Todo

1. OpenAI-ES vs Gymnasium

   1. Lunar Lander

   2. Bipedal Walker

   3. Car Racing

2. OpenAI-ES vs my own environments

   1. Snake

   2. Flappy bird

3. OpenAI-ES in an intro to multi-agent systems (design my own environments)

   1. Tron-legacy light cycle chase: 1v1

   2. Backgammon

   3. Checkers

   4. Card games

   5. Tron-legacy light cycle chase: 2v1

4. Implement NEAT & HyperNEAT algorithms, and use them for the same previous games

5. Packaging

   1. OpenAI-ES, NEAT & HyperNEAT in an abstraction similar to PyGamo

   2. Add a suit of easy benchmark problems (critical to develop new algorithms)

Done

1. Implement OpenAI-ES algorithm: since it is pretty simple, and works well with neural networks

To Explore

1. Grid environments: whether to use MiniGrid or create my own, and what kind of questions / environments to investigate

2. PettingZoo for multi-agent environments

3. Experiments from "Why Greatness can't be planned"

4. Interesting games

   1. Tower defense

Phase 2: TBH

Objectives

1. Work with networks (networkxx)