I would love to solicit suggestions for other ML topics and languages to cover as part of this series (please don't suggest "SVMs in brainfuck"). Let me know what you think!
I really liked the way you have presented the topic in the article.
One suggestion I have is to just stick with one language instead of using multiple languages in the series. I understand that you are using different languages to prove the point that ML can be implemented in any language, but when you switch the language between articles, it might get a little awkward to follow. I would suggest teaching in one language which you think can better represent the logic and then provide a link to a github repo or something which has the same code in different languages.
I started a little genetic algorithm project that I haven't gotten to work on too much yet, which could make for a cool tutorial. I've got a GA that solves the traveling salesman problem in the Go programming language. It actually solves it in N isolated populations, where you tell it to use N go routines. (Optimally I believe N would be 2x the number of cores your machine has).
Since Go is a systems language, I have it running as a http server that sends back the results to webpage that renders the JSON on a HTML5 Canvas.
What I really wanted to do is experiment with "migrating" solutions from one population to another, to see if I can get any speedup that way. Right now, the isolated populations sort of converge at the same speed, but some do better than others, because they converge to different local minima.
If you're going that route, I suggest instead looking at Grammatical Evolution, which is much more popular and does a very closely related thing. Or heck, stick with plain Genetic Programming (GP).
It'd only be a toy though. Because of the high number of evaluations used by these kinds of methods, and the cost of a single evaluation, most serious work is done in engines based on faster languages, notably Java, C, or, or C++.
