Revision of Episode 2 (Using your GPU with CuPy)

[wmotion]

The revision in this pull request covers specifically episode two. At the time of writing, it did not cover the questions and objectives (top of page) and the keypoints (bottom of page). The last three sections have been revised only lightly; suggestions as to how to proceed are given. Specific issues may follow at a later time.

The topic branch tied to this pull request is Episode-02 (thus, not main). Its first commit is 'empty' (that is, no change in working tree) just to mark the detachment from the main branch. This branch also contains merges from secondary branches dedicated to code changes, linguistic changes, pedagogic changes. The arrangement of these secondary branches in the fork is my own experimentation with content management: they should be of little import for the pull request.

The beginning of the commit subject indicates whether the focus was on code changes, linguistic changes, or pedagogical changes. The last two are largely tied to each other: pedagogical changes implies more attention for the process of learning and for ways to encourage learning and retention. Some commit messages are lengthy with a (hoped-for) purpose to explain the underlying assumptions and possible alternatives.

Of course, it holds for all edits that the course developers/content experts should double-check whether the new text has stayed true to the original intentions. Long live version control!

SPECIAL COMMITS

The commits from 1abb082 through cdf743d regard semantic newlines (one sentence = one line) and rote spellchecking. This handling makes it possible exploit the Git toolbox to maximum clarity, using the same units Git uses for comparisons. Think of informative git diff and git merge operations. Therefore, pulling these commits should be fail-safe and is recommended at any rate.

Commits db44928 and 591d289 explain the editing approach to the first half of the episode (Convolution), at least as is was conceived at the start. Other commits could also contain relevant annotations, though.

Commit 9dd5c8c explains the unfinished work on the second half of the episode (image processing in astronomy).

ACKNOWLEDGED ERRORS (SO FAR)

In commit 9dd5c8c I had not understood the direction of the image processing method yet. The standard deviation of a clipped data is used as standard deviation for the entire data set (as I understood it now). Please take this as a piece of evidence that the method is not explained clearly.