Thank you (again) for being interested in our work. I enjoyed your video very much! Especially the comparisons at around 44:00 and 47:30! I hope you won't mind if I include some of these comparisons into my slides for the paper's presentation in the future :) I just want to clarify one thing (at about 09:00), we chose ES because we happen to have an ES training codebase that served us quite well in the past. As we mentioned in the paper, AttentionNeuron is differentiable, so it is possible to training using gradient based methods too. In fact, we trained Ant and Pong using behavior cloning in the paper. Again, I love this video and would like to thank you for the effort!
great work indeed so far !, If you consider, I would like to give a suggestion for your next videos, if you put the equations of methods alongside with your text editor, e.g. splitting up your screen into two pane and then slide down equations while writing your code. It makes easier to follow up the algorithm for the audience, and definitely increase intuition.
That is indeed a great suggestion ❤️. Thank you! I am always worried about making the code too small to read, however, yeh, the right side of the screen can definitely contain some diagrams/equations!
Thank you (again) for being interested in our work.
I enjoyed your video very much! Especially the comparisons at around 44:00 and 47:30!
I hope you won't mind if I include some of these comparisons into my slides for the paper's presentation in the future :)
I just want to clarify one thing (at about 09:00), we chose ES because we happen to have an ES training codebase that served us quite well in the past.
As we mentioned in the paper, AttentionNeuron is differentiable, so it is possible to training using gradient based methods too.
In fact, we trained Ant and Pong using behavior cloning in the paper.
Again, I love this video and would like to thank you for the effort!
I appreciate your comment:) Sure, feel free to use anything from the video/code!
Thanks for the amazing walkthrough and implementation! This was awesome :)
Thank you!
Time to enjoy a new machine learning adventure with this video 🎉
Good job! Loved it! ❤️
insanely helpful, as always. many thxs
I am the first one to give this video a thumb up.
great work indeed so far !, If you consider, I would like to give a suggestion for your next videos, if you put the equations of methods alongside with your text editor, e.g. splitting up your screen into two pane and then slide down equations while writing your code. It makes easier to follow up the algorithm for the audience, and definitely increase intuition.
That is indeed a great suggestion ❤️. Thank you! I am always worried about making the code too small to read, however, yeh, the right side of the screen can definitely contain some diagrams/equations!
Quality content 👌🏼
Do you have the code for pong implementation?! Or how they create the random patches view. Thanks for sharing this
Sorry, I did not look into the pong task at all.
Wau. Another great video:)
👍👍👍
This is so complicated to undesrtand, :(
Sorry to hear that. What part are you struggling with?