>>12374750If you just wanna learn how to design NN achitectures it’s best to do Stanford’s CS231n, read the papers and get implementing. This is a pretty pleb route though will get you up and running quick
Proper ML theory can get VERY demanding mathematically. I’m assuming you have pre-reason like Undergrad Linear Algebra, Calculus and stats covered.
The most rewarding (but also time consuming) path to doing this is to read Vershynin‘s “High Dimensional Probability” cover to cover (trust me this shit will take time)
https://www.math.uci.edu/~rvershyn/papers/HDP-book/HDP-book.html#Then you could move on to Shai and Shai’s text
https://www.cs.huji.ac.il/~shais/UnderstandingMachineLearning/understanding-machine-learning-theory-algorithms.pdfA moderate direction to take would probably be Bishop’s Machine Learning text (or Kevin Murphy)
I strongly recommend the Vershynin then Shai route if you have the time and dedication for it but Bishop should be fine otherwise
