During the FruitPunch AI Hackaday, teams will take on a number of AI challenges. These Challenges will be presented in the form of Classic Atari games. The goal for the teams is to program and train an artificial intelligence that will play the game as best as possible. They will receive the allotted time during the event to do this, with preparation beforehand.
🙋♂Teams can consist of 3 to 5 people.
🍉 The day starts at 09:00 when we will welcome the teams and explain the exact challenge. After setting up their hardware, the first timeslot for working starts at 10:15.
🥝 At 14:00 will serve lunch and at 19:00 dinner because human intelligence literally needs ‘food for thought’.
🍒 The second timeslot to work on their code starts at 14:30, this timeslot ends at 18:45.
🏆 When we’re done digging through your code the winner will be announced at 21:00, after which we’ll have a celebratory drink. It will be very exciting to see the fruits of everyone’s labour, and which ones taste the sweetest.
🎟Team registration is available through the form on the right! 👬
Good luck, and have fun!
Date: 30-03-2019
Time: 09:00 – 22:00
Location: TU/e, Auditorium
Teams: 3-5 people

Prior knowledge and preparation

On the Hackaday itself you will need to make use of the Pycman package made by Team Serpentine.

To keep things fair you can download the package here to get familiar with it.

The zip file contains Pycman, some instructions for booting the package and a small manual. It is advised to take a look at it beforehand, since this is also the package used on the hackaday itself.

(ps. read the readme ;p)

Pycman package download:

To really shine on the Hackaday it is advised to have or do up some knowledge on reinforcement learning and how to build a neural net. A basic version of an AI will be given and how high you score will be based on how much you can improve it. The Atari games will be run in OpenAI Gym and algorithms will be implemented using the Pycman package. If you missed the masterclass, check out the following tutorial:

Jupiter notebook tutorial reinforcement learning:

We’ve also concocted a document containing helpful tutorials that we thought will be usefull in preparation for the Hackaday.

Prerequisite knowledge tutorials: