TangrAI

One Training Episode

Inspiration

We are machine learning lovers and we wanted to perform a new project around reinforcement learning.

What it does

Our program has learnt from zero, how to resolve the Tangram puzzle. The game starts with an empty table, and it have 7 pieces available for resolving the puzzle. The program has learnt itself which order it will put the pieces and the position in where goes each piece.

How we built it

A reinforcement learning project has three main parts: the environment, the agent and the engine. We have built all of this parts from zero, because it has no other projects for Tangram Puzzle. Although, we have used scripts from Github for having an idea.

Challenges we ran into

Carles is a electronic engineer and Pau is a mechanical engineer. Our main challenge has been the performance of a computer science project of huge difficulty without have studied any degree in computer science. For us, that we love machine learning, this means that we have enough capabilities for achieve and solve successfully computer science projects.

Accomplishments that we're proud of

We are really proud of finishing it during this hackathon, due to we thought that we would finish our project at home on the next weeks. However, we have not had enough time for finishing the training because it takes to much time and we have run out of it.