Snake Game, Controlled by either a human or AI player

This project is the classic snake game, and it can be controlled either by a human or an AI.

The AI part is using a feedforward neural network to train its game play.

Here are some links to the theoretical aspects of the AI.

PyTorch Module

The game is using the PyTorch library, and its Module base class. Read more about it here:

https://pytorch.org/docs/stable/generated/torch.nn.Module.html

Activation Function

We are using the ReLU activator function to transform the summed weighted input. This article describes ReLU.

https://machinelearningmastery.com/rectified-linear-activation-function-for-deep-learning-neural-networks/

Optimizer

As our optimizer we are using ADAM

Here is an article about why we use optimizers https://www.geeksforgeeks.org/optimization-techniques-for-gradient-descent/

There is also a nice video describing them here https://www.youtube.com/watch?v=mdKjMPmcWjY

To get an understanding of what the ADAM optimizer does, you can go here https://www.geeksforgeeks.org/intuition-of-adam-optimizer/#:~:text=Adam%20optimizer%20involves%20a%20combination,minima%20in%20a%20faster%20pace.

Loss Function

We use MSELoss, to measure the loss at each learning step. Read more about it here https://pythonguides.com/pytorch-mseloss/

Learning Rate

Read this article to get an understanding the impact the learning rate has on Nural Networks https://machinelearningmastery.com/understand-the-dynamics-of-learning-rate-on-deep-learning-neural-networks/

Backpropagation

To understand the role of backpropagation you can look at this Wikipedia article https://en.wikipedia.org/wiki/Backpropagation