(Day 123] Optimization algorithms chapter from Dive into DL

Ivan Ivanov · May 3, 2024

Hello :) Today is Day 123!

A quick summary of today:

  • read about optimization algorithms, and maths for DL from Dive into Deep Learning

Firstly, my notes from the optimization algorithms chapter

Covered topics:

Common challenges in DL optimization, convexity, convexity properties, SGD, momentum, Adagrad, RMSProp, Adadelta, Adam, LR schedulers

image image image image image image

As for the maths for DL

This is actually in the apendix, but I decided to read through it, just to check if there is any interesting math.

The content is as follows:

image

Having this appendix is very nice and definitely helps to answer questions like for example when I went to the roots of multicollinearity (on Day 121) - understanding the math helps understand the root of problems.

As for the MLx Fundamentals it started just as I am finishing this blog. The first day is from 9pm to 3am in my local time, so I will write about it in tomorrow’s blog.

Schedule (time is in BST)

image

That is all for today!

See you tomorrow :)