Do research

[Ablesh1]

1. Selection of ML algorithms to be used in the project
2. Literature review of ML algorithms
3. Code review of ML algorithms

https://learningtoplaydotnet.files.wordpress.com/2020/08/ptl4.pdf

foundations-of-deep-reinforcement-learning-theory-and-practice-in-python-1nbsped-9780135172384-0135172381_compress.pdf

[Krzychu-Z]

First possibility for our ML algorithm: Deep Q Learning using Q-values table. Below explanatory video: https://www.youtube.com/watch?v=t3fbETsIBCY

In this scenario our NN learns from movement sequence "sadawdwasdasdqwdwdasd..." (string or char array) serialised into game engine. TBD: length of this sequence (longest possible gameplay?)

[LeMeTwo]

Take my suggestions with a grain of salt - I am not sure how to app[y these algorythms with our app exactly, but those are quitre popular and well researched.

As @Krzychu-Z mentioned Q-learning I suggest comapring it to Deep Q-Network (DQN) https://towardsdatascience.com/deep-q-learning-tutorial-mindqn-2a4c855abffc

Another one, for a game can be actor-critic https://www.researchgate.net/figure/The-actor-critic-proximal-policy-optimization-Actor-Critic-PPO-algorithm-process_fig3_339651408

If we decide to use A-C, we can then compare it with Proximal Policy Optimization (PPO), as it`s similar more or less, yet oftentimes offers better results. https://towardsdatascience.com/proximal-policy-optimization-ppo-explained-abed1952457b