News for hackers. Stuff that matters.

diegoalejogm

diegoalejogm on Hacker News

Re:Adversarial Network Learning MNIST Digits on Thursday February 01, 2018 @10:29AM (#16282485)
Make sure to expand the VGAN_MNIST tab to see error progress in real-time.

Also make sure to click on the IMAGES tab to visualize the progress. There's a slider to see the images generated at previous steps, and the current ones!!

ENJOY

Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @08:09PM (#16236265)
Good to know :) Thanks for the info, I'll have that in mind for the following posts!
Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @06:48PM (#16235859)
Forgot to mention that it is learning from raw pixels!! No other info is passed to the agent, except the screen frames :)
Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @05:14PM (#16235211)
No worries! Here's also a link to the current progress https://www.floydhub.com/diegoalejogm/projects/atari/29 in case you wanna check the machine usage too
Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @05:12PM (#16235197)
Yeah! I couldn't have described it better than you just did!

Feel free to take a look at the code and star the code if you like it. I'll keep upgrading it and adding more Reinforcement Learning variations

Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @03:49PM (#16234532)
No, it hasn't. Here: http://github.com/diegoalejogm/deep-q-learning

Try again and let me know if this time it works :)

Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @03:47PM (#16234511)
Also, it will be training for several hours more (I approximate 20), so make sure to keep updated. I'll post videos on the current execution performance.
Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @03:44PM (#16234480)
Also, you can check the FloydExecution in: https://www.floydhub.com/diegoalejogm/projects/atari/29
Re:AI Learning How to Play Pong in RealTime – PyTorch on Thursday January 25, 2018 @02:44PM (#16233910)
PyTorch code: https://github.com/diegoalejogm/deep-q-learning

I'll add final results and performance videos soon!

Re:AI Learning How to Play Pong FloydHub and GitHub Links on Thursday January 25, 2018 @02:22PM (#16233692)
I recommend to soften the curves with a value of 0.9

Here's the link of the FloydHub process: https://www.floydhub.com/diegoalejogm/projects/atari/29

Re:Deep Q Network Learning to Play Pong on TensorBoard on Tuesday January 23, 2018 @01:46PM (#16216171)
Yeah! Sorry for that, FloydHub has got some errors at the moment and my progress wasn't stored anymore.

You can check out the code here: https://github.com/diegoalejogm/deep-q-learning

It works! I got the models up to the point where FloydHub stopped working. Leave me your mail or something and I can write you to let you know when I upload the training model again! :)

Re:Deep Q Network Learning to Play Pong on TensorBoard on Tuesday January 23, 2018 @12:27AM (#16211251)
I built it with PyTorch and is running in FloydHub with GPUs. I suggest you set a smoothing of 0.9 to see the increasing trend :)

Code can be found here: https://github.com/diegoalejogm/deep-q-learning

Re:Generative Adversarial Networks Code in PyTorch and Tensorflow on Friday January 05, 2018 @06:56PM (#16082836)
It looks awesome, thanks!

The idea of this library is to show the inner workings of the GANs as they’re publicized in the papers, but I will definitely make some examples with TFGANs in the future for broader exposition. ;)

Re:Generative Adversarial Networks Code in PyTorch and Tensorflow on Friday January 05, 2018 @05:30PM (#16082338)
That’s the next step, after implementing more GAN papers. If you’re up to it, you’re very welcome to contribute with that. :)
Re:Generative Adversarial Networks Code in PyTorch and Tensorflow on Friday January 05, 2018 @02:13PM (#16080823)
More models coming soon! :)