Notes on “Meta Learning - How To Learn Deep Learning And Thrive In The Digital World”

I recently bought this short book Meta Learning: How To Learn Deep Learning And Thrive In The Digital World by Radek Osmulski.

It’s only 90 pages, available as mobi or pdf.

Here is the summary:

Meta Learning is an actionable roadmap to learning machine learning efficiently. It will show you exactly what you need to learn and how to learn it in order to become a world-class machine learning professional in the least amount of time.

Below are some of my notes from the book.

1. Familarise yourself with an editor (e.g. VSCode, Emacs, VIM etc.)
2. Familarise yourself with Git
3. Go half-half between theory and hands-on. Don’t be too obsessed with theory and leave out hands-on
4. Try to get focus-time and avoid distractions
5. Start something small and share with people
6. Establish a baseline to determine if your subsequent efforts are moving in the correct direction
7. Start small by training on subset of data (caveat is some models dont train well with small amount of data)

Some good links:

1. The Missing Semester of Your CS Education
2. How (and why) to create a good validation set
3. Kaggle Ensembling Guide
4. Kaggle Ensembling Guide Repo