Is there a way of determining how many layers of a neural network are neccessary for it to abstract/generalize a concept? E.g. if we play chess: Obviously as you sayed one layer would not be enough. But can we know or proof how many layers I need such that the network develops concepts such as control of the center or block a passed pawn etc?
+gcgrabodan There's rarely an exact number of layers implied by a problem, and the details tend to really matter---more layers can sometimes even make things worse. For now anyway, neural net design is still as much art as science.
Your work is super interesting, but I would like to give a huge positive cookie for your presentation style too. I have watched a lot of your videos by now and besides the actual topics that you cover, I have learnt a lot in how to present ideas in a calm, meaningful, reserved but still super interesting and fun way. You have a new... fan? Something. Best of luck with your work Dave!
Aw man, your vids are so cool and easy to understand! It's sad that I've only discovered this channel just now!
Some of the best compsci vids I've ever seen, thank you so much! :)
Alex Conway Thanks for the comment!
you are a tremendous teacher...this channel's gonna be big
Mattijs Van De Walle We'll see! Thanks for the comment.
Wow, this guy actually looks and sounds more natural at a speed of 1.25! Love this topic and this video!! Thanks dude
Wow an awesome explanation you are good positive reinforcement ok no
😅
Excellent, all of these vids are excellent man, thanks!
Is there a way of determining how many layers of a neural network are neccessary for it to abstract/generalize a concept? E.g. if we play chess: Obviously as you sayed one layer would not be enough. But can we know or proof how many layers I need such that the network develops concepts such as control of the center or block a passed pawn etc?
+gcgrabodan There's rarely an exact number of layers implied by a problem, and the details tend to really matter---more layers can sometimes even make things worse. For now anyway, neural net design is still as much art as science.
what does the y axis of the beta distribution graph represent?
Your work is super interesting, but I would like to give a huge positive cookie for your presentation style too. I have watched a lot of your videos by now and besides the actual topics that you cover, I have learnt a lot in how to present ideas in a calm, meaningful, reserved but still super interesting and fun way. You have a new... fan? Something. Best of luck with your work Dave!
Yum yum tasty! Thanks for the note.
Great learning video, thanks for your effort :)
You sir, are amazing. Great Video! :)
Amazing stuff. Very well explained.
Fun video, I enjoyed this a lot!
Can we talk personally...?
So what you saying is..... Bush is a cat robot, got it! :D
ads will get nothing but negative cookies from me >:C
+Estlib Anon The Breakfast of Champions! Oh wait.
We not going to just gloss over that shroom quip.
cool presentation! i really wish you were my teacher...
anyone else think of plumbus when he said fleeb and gloob?