AI Plays Pong - Reinforcement Learning

• 

  Pong AI

Inspiration

I saw the original Pong game and recreated it. I had no one else to play against, so I decided to take the project to a new level and train an AI to play.

What it does

It involves having the AI play the game and learn how to score against the opponent by collecting data and using it to train a model in Tensorflow.

How I built it

I used Pygame to set up the reinforcement learning environment and saved data from the game into a csv file. From there, I set up the process for training the model. Finally, I deployed the model into the game to play against a human player.

Challenges I ran into

One challenge was that the model was not predicting correctly and the AI continuously made incorrect movements.

Accomplishments that I'm proud of

I managed to overcome the training issues mentioned above. I rewrote the method in which I collected the data.

What I learned

I improved my programming skills in machine learning by working on optimizing the game.

What's next for AI Plays Pong - Reinforcement Learning

I would like to improve real-time machine learning so that the AI can learn quickly as it plays the game

Built With

• keras
• pandas
• pygame
• python
• tensorflow