It states that deadline is 9.13.2020b
Would be nice to correct it.

Expand Adaboost notes with sections for bonus points:

• Adaboost derivation, 2bp handnotes
• Adaboost proof, 2bp, handnotes + proof
• Cooler demo 1-2bp
• Viola Jones paper w 1bp face detector

There was a link to doodle about the term of the final exam. It would be nice to have it somewhere in the repo.

In this year's "Estimate performance for various ks" section, the method for preparing the training and testing datasets has changed compared to the last edition - from plain 2:1 split of the dataset to bootstrapping error estimation.

Last year sample code looked like this:

Now:

I spent some time thinking how np.split could be useful in this situation, but couldn't come up with any sensibile ideas.
Thus I think, however I might as well be just biased by my own solution, that the train_idx, test_idx = np.split(...) is not right tip for a solution.

Note: I decided not to create a pull request as suggested during the first lecture. Firstly, I am not sure if I am right. Secondly, I wouldn't want to spoil the solution.