machine_learning

Here is a backstage from this course that may be useful for people who are interested in teaching or presentation.

I found out that Colab platform is cool as it allows to merge executable code, nice formulae formatting, pictures, etc. I believe it’s a new level of presenting lectures to the students. Github allows all materials to be readily available any time the student wants them. It also eliminates the need of taking notes. A student can focus on making sense of the material he is exposed to and can return to the exact same points he is told at the class any time later. However colab notebook has features of a book and a powerpoint spreadsheet so it requires a specific presentation (with some material scipping and handwaving) in order to keep the lectures swift. I used a laptop (with screen projected onto the wall) with the drawing on the screen feature in order to highlight details. I also used the blackboard for sideline explanations.

The backbone of the course are the python examples covered in the Python Data Science Handbook by Jake Vanderplas. However I realized I needed to add

• python introduction (1.5 lectures);
• matplotlib, pandas, and numpy introductions extended with practical problems in workbooks (3 lectures);
• statistics and mathematical optimization overview (2 lectures);
• extensive theory injections in each lecture;
• interactive examples;
• seminar tasks.

I tried to watch and compile as many online youtube lectures or stand alone videos as I could and tried to give credit to all of them in the Further Resources section of each lecture.

I also realized that scilearn documentation is very helpful and should be extensively used. There are references to actual scientific papers that scilearn algorithms are based on. If you want to go in depth you should check them out.

Adding interactive widgets is very important as it helps to develop intuition into what is going on in the lecture. Playing around with points is very engaging and fun for students. This widgets should be seen as laboratory device that students use to gain understanding on the “laws of nature” (as is done in physics classes).

I tried to inject as many multimedia as I could including:

• pictures (drawn in Inkscape);
• interdisciplinary references (biology, physics, math, you name it…);
• pop culture references that are relevant to the topics discussed;
• datasets of historical value (John Snow cholera dataset, Pearson’s son/fathers height dataset, Old Faithful dataset).

I tried to inject humour and gags accidentally in order to keep students amused and lighthearted. I am not sure how that played out though:).