Here's another case where I might have stumbled on an andragogic technique another Python teacher is already well-known for using. Or not, we shall see. Old technique (for teaching properties): In an earlier chapter, I stumbled upon having @property decorate a circle so you could change radius, area or circumference with simple "setattr" dot notation e.g. c.area = 10, and the other two attributes would change automatically. [1] Thanks to how type property uses the Descriptor pattern, that's quite doable and is a clear demonstration of what properties allow, drawn from familiar grade school geometry. Turns out: Raymond Hettinger was sharing that little dharma already. Same karma! Great minds think alike (if I do say so myself). Aside: I've been meaning to do more with triangles and tetrahedrons, using @property... e.g. make AB longer and watch angles change. Sticking to right and/or equi-angular triangles keeps everything simpler. [2] The new technique (for introducing data structures and machine learning): So the new thing I might not be first to think of: lets use the "chopsticks pattern" from the musical score of Chopsticks (used universally in tutorials, almost a "hello, world" of Piano World) and call the chopstick note pairs "correct" amidst a myriad "not correct" bytes. Here's an octave: C D E F G A B C Chopsticks begins with F and G pressed. Then E and G. Then D and B. Then C and C. The first 22 seconds of this Youtube give the idea: https://youtu.be/waraNMP0kK8 So in "byte format": 0 0 0 1 1 0 0 0 0 0 1 0 1 0 0 0 0 1 0 0 0 0 1 0 1 0 0 0 0 0 0 1 are the "chopsticks" of interest. Then have Machine Learning algorithms tease out the pattern. Feed through 10,000 random strings of 1 and 0. [3] Mark the chopstick patterns "correct" (1) and the not-chopstick patterns "incorrect" (0), effectively forming a ninth column (the proverbial y in machine learning, where all the samples are X). This way, we get to play with (introduce) numpy.ndarrays and scikit-learn, but with more familiar thoughts about piano keys in the foreground, and a melody to boot. How good are these learning machines, once trained? Do they get random 10010100 right i.e. "not a chopstick"? Are they right every time? If intrigued and want more code, here's the link to the Jupyter Notebook in question: https://github.com/4dsolutions/SAISOFT/blob/master/OrderingData.ipynb (scroll to very end and come backwards would be my suggestion -- get the ML part first). I like how something so early in piano training feeds an intro to ML, given how piano and "player piano" relate to AI, of which ML is a part. Punch cards and all that. Very Westworld eh? https://youtu.be/elkHuRROPfk (not just Chopsticks anymore) I'm looking for "pathways through Python" that consist of a combination of "zoomed in" and "zoomed out" topics. Sometimes we look at nitty gritty, other times we need overview. Kirby [1] this older version (Oct 2016) doesn't have perimeter (circumference). Easy to add? (we do that as an exercise in class). https://github.com/4dsolutions/Python5/blob/master/Descriptors%20and%20Prope... [2] I've got this dynamite volume method, not invented by me, that just takes the six edge lengths for the arguments. https://github.com/4dsolutions/Python5/blob/master/tetravolume.py (used a lot in my stash) Lots more in the historical literature. E.g.: https://www.mathpages.com/home/kmath424/kmath424.htm More context: https://medium.com/@kirbyurner/uncommon-core-87a31b7f75b3 [3] My current function for doing that is maybe too long-winded as I concatenate strings. Why not just convert to binary from random 0-255. We could do that.