Visual prerequisites for learning deep learning
Every time I keep repeating myself I feel the urge to write things down so that I can simply point such resource in future occasions. Every semester I start my course showing a collection of visual resources for priming my students’ mental picture muscle. By the end of the semester, they are usually able to see a little more.
Everything I teach
Previous edition of my courses can be found under didactics. I teach from primary and middle schoolers, to PhD students, to retired folks. From maths, science, and programming to music, dancing, and cooking.
From my (former) high-school intern
Vivek put together these two manims for me:
Searching Twitter
To know my (or others’) thoughts on SVD search for (from:alfcnz) SVD
on Twitter.
Other examples:
(from:alfcnz) linear algebra
(from:alfcnz) probability data science
If you become a follower, you’ll be automatically notified with new learning visual content.
Maths
Is your maths a little rusty? No problem. Check out Marc’s book Mathematics for Machine Learning. You really want to read Chapter 5, Vector Calculus, before homework 1.
Linear algebra
Please, binge watch Essence of Linear Algebra by Grant. This is a requirement.
For a more in-depth coverage of the topic, I like Gilbert’s Introduction to linear algebra and corresponding video lectures.
NEW «Linear Algebra and Its Applications» by David & Steven Lay, and Judi McDonald.
«Linear Algebra and Its Applications» by David & Steven Lay, and Judi McDonald is truly a masterpiece.Load tweet
A textbook teaching Linear Algebra geometrically, application first, with computer exercises and data sets, interactive figures, and projects.
🤩🤩🤩 pic.twitter.com/kQvoTuHHvP
Trefor
His YouTube channel is worth your subscription. Full playlists for discrete maths, linear algebra, calculus I-IV, and differential equations.
Eigensteve
Do you need to brush up control? Check out Control Bootcamp. Additional playlists include differential equations and dynamical systems, complex analysis, vector calculus and partial differential equations, and singular value decomposition. Steve’s YouTube channel is full of extremely well done lectures.
Blogs
Shameless plug
Do eigenvectors need to be orthogonal?
(You keep giving me the wrong answer. Yann included…)
May be worth checking out Eigen-stuff.
Gregory
His blog is stellar. I recommend particulary:
Network drawing software
Draw.io is what I use to draw my slides’ diagrams with. It could be a useful tool for you too.
TikZ
Yet, for my textbook I’m using $\TeX$-based TikZ. Why? Because I’m crazy. Another aspect is that, being text, you can edit it any time, apply whatever theme you want, and the maths in the figures will be updated if I change any definition in the preamble. Moreover, I can draw parametrised figures, where lengths and sizes are the result of computations.
Similarly, all plots are also $\TeX$ generated with PGFPlots. Yes, it’s a bit painful at the beginning, but the final quality is extremely satisfying. In fact, all plots in my textbook are drafted with Matplotlib and then recreated with PGFPlots, which reads some ASCII files, where I’ve dumped the data points.