We saw a video of Sethbling demonstrating a similar project that made Mario learn from his mistakes to eventually perfect his run.
The Lua script that we wrote gets data from the emulator, which uses it to self-learn to play itself. We used genetic algorithms to assist the machine in getting farther in the level of the Mario game.
We first got inspiration from Youtube videos. Then we picked an idea and started brainstorming the algorithm. We later wrote the code and started debugging it.
The biggest challenge was creating the cross algorithm that made the next attempt, after failing, have greater success.
There were a lot of problems that we faced, especially since Machine-Learning was a totally new thing to us. The language Lua was also new. After struggling for the day, were proud that we held our high hopes and ambitions, and passed through the struggles.
A new friend that we met was kind enough to gives a quick introduction to what ML is and what it can do. We also learned a lot of new techniques that we can used in the future when facing something new. Although it was daunting at first, we are know more confident.The experience of struggling for the day, really did pay off!
We hope to make the algorithm far more efficient than it already is. This will bring unprecedented challenges that will be both daunting and exciting.