Hello everyone,
I completed the maths spec and I just wanted to share these few tips, hopefully it helps someone.
Here are the things that worked for me:

Psychology: revisiting math as an adult is actually way easier than it was as a kid. First you took this course by yourself, instead of being forced to learn, you are coming from a growth mindset, remember that. Secondly you have more context and maturity as an adult and finally you got an objective: apply the math to machine learning.

What is a formula anyway? When you come across a formula, there is no reasonable reason to assume that you would understand it from the first try. I mean, really, it’s not about being smart enough; it’s just a matter of perspective. Take a step back and look at the bigger picture. Remember, every formula is the result of someone’s hard work and dedication. They observed, experimented, and tested countless times to find the perfect way to express a truth about the world.

Plot everything: So a formula is a formalization of some truth, it’s like a fact, it’s dead! To understand the fact you need data to bring it back to life, you need to experiment and observe. The good news is that we live in 2024, you just need to plot something for 10000 times and look at the graph. Everything start to make sense with scale!

Greek symbols: Don’t get afraid of them, there are not so many!

Code is truth: What seems scary on paper, is actually very manageable once translated in code, don’t get impressed, also the power of code is that you can get feedback while paper is dead!

Usual suspect number one: We often use square to turn negatives into positives, and since we use this trick, we sometimes need to translate that value back, this is done through square root. What’s the difference with abs then? Squaring stretches values, this can be useful to make outliers more obvious.

Usual suspect number two: You will often see logarithms, it’s very often because of mathematical tricks of adding logarithms, that’s it. Another useful aspect about log is that it compresses ranges of data, remember when we said that square make outliers more obvious? Well logarithms do the opposite.

Usual suspect number three: A matrix is really just an excel sheet, imagine a table with different columns and where every row is a new entry. You can also see every row as a linear equation. Matrix multiplications are like blueprint for going from dimension A to B, there are multiple ways to remember what cell to multiply with which cell, just find a trick that works for you and stick to it.

Mathematical Gymnastics: When you think about it, there’s not much things to do actually, we turn numbers into positives (squares), then we bring them back to normal (√), we make sum (Σ), we make averages or normalize (/n). We compress (logs) or we stretch (exp), etc. There are all kind of gymnastics to manipulate data in a certain way, but we very often encounter the same tricks.
Your only job is to not give up, good luck!
I would like to send my thanks to the deeplearning.ai team for making such a great course!