In this problem, we using a novel deep reinforcement learning technique to achieve state-of-the-art results in the Atari game: Kangaroo. Out current plan is to release the results of all Atari games in January 2019, but for now…We leave you with this preview :)

Following the standard practice in the Atari domain, we need to report the best snapshot from a 200 evaluation episode rolling average. Below you will find our preliminary plot for the Atari game: Kangaroo, that corresponds to the video that can be found here YouTube. If you have any questions, feel free to contact me by email!

Results:

State-Of-The-Art: 14,854.0

Novel Approach: 15,780.0

Kangaroo