I just know and reading this book: Emergence: From Chaos To Order (https://www.amazon.com/Emergence-Chaos-Order-Helix-Books/dp/0738201421),<p>but it just studied simple shadow neural network,<p>Are there any research using chaos theory to study Deep neural network?
A lot of it you can check here <a href="https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=chaos+deep+learning&oq=chaos+dee" rel="nofollow">https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=chao...</a> or <a href="https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=chaos+theory+deep+learning&btnG=" rel="nofollow">https://scholar.google.com/scholar?hl=en&as_sdt=0%2C5&q=chao...</a><p>Also try to search for chaos in reinforcement learning. Hope it will help
Chaotic systems are iterative, that is, you keep data through the same function over and over again.<p>As usually used, neural networks run in a single pass, you get data in and get data out.<p>If you iterated a deep neural network you'd probably find chaos and other dynamic phenomena. But if you just run it once, nope.