Deeplearning4j - Reinforcement Learning for Java - Deep Q Learning with double DQN - Snake Game

In this video we take a look at Deeplearning4j and try to implement a neural network that can be trained to play a game of Snake. Algorithm that we are using is Deep Reinforcement Learning with Double Q-learning. More info about it can be found in the links below.
Source code:
github.com/lil...
Useful links:
deeplearning4j...
arxiv.org/abs/...
Intro by Audionautix is licensed under a Creative Commons Attribution 4.0 licence. creativecommon...
Artist: audionautix.com/

[@mohsengh6415]

Great video, thanks

[@Michael80B]

Thanks for sharing this. It's something I've been looking for a long time. I got your code running and wanted to do some experiments without having to wait too long for feedback. So I turned off all the UI and hoped to get learning within seconds. After all it takes 1 - 5 seconds for each game!?!
Is the speed that bad because the app not optimized? Is it because Java is not the best tool for RL? Or do one just to be patient when letting the computer learn?

[@koxi7709]

Thank you

[@hayk.avetisyan]

thanks for video

[@sergiveramartinez2685]

Hello! First of all, great video.
I was trying to import the project in IntelliJ but it doesn't work, since it's a gradle project. Do you have a video explaining the environment configuration? Many thanks

[@liliumbosniacumcode]

Thank you. What kind of errors do you get ? Maybe you can send me an email explaining what the issue is and I can try to help you out. Email is on the about us page on the channel.

[@sergiveramartinez2685]

@@liliumbosniacumcode Thanks Lilium!! I’ll send you a message👍

[@user-zk6kq6yt1j]

Hello, I would like to know about how to set the range of actionIndex which output from network. Actully, the lenght of data array that I give to network is so long, so it will not equal to the range of output actionIndex. I had tried to change the return value of GameObservationSpace.getShape(), because I have attation seemingly the actionIndex in the range of the input size. Therefore, I would like to know how to set it.