search

nature-of-code / NOC-S17-2-Intelligence-Learning

Syllabus for Part 2 of Nature of Code: "Intelligence and Learning" at ITP Spring 2017 Edit
934 stars 341 forks source link

Reinforcement Learning comes too late? #6

Closed shiffman closed 7 years ago

shiffman commented 7 years ago

Better to do earlier somehow due to connection to steering behaviors?

cmanny commented 7 years ago

Actually, I think you are right here, I would agree that i should come after Evolutionary methods

shiffman commented 7 years ago

Yes, but wouldn't it depend on covering neural networks first?

oxxie commented 7 years ago

Actually, I would consider reinforcement learning as a good extension of neural networks, so I'd suggest it should come directly after week 4. Maybe switch weeks 5 and 6? Convolution is a much 'broader' topic, am I right?

shiffman commented 7 years ago

@oxxie this is a good point, I need to get a little further in preparing examples / picking frameworks, but I agree that this could be a good switch.

oxxie commented 7 years ago

Glad to help ;-) By the way: I have been preparing a matrix calculations library for Processing, which might ease the neural networks part a bit, e.g. you don't have to write your own code for transposing and Hadamart products etc.. Drop me a line if you are interested, would love to share.

shiffman commented 7 years ago

Going to close this now, I failed to do a good job of fitting reinforcement learning in, but I hope that it will be a primary subject of the new nature of code chapter 10.