Learning to Play Tic-Tac-Toe with Jax
joe-antognini.github.ioYou can build a machine that plays perfect tic-tac-toe with 300 matchboxes.
https://en.wikipedia.org/wiki/Matchbox_Educable_Noughts_and_...
The first time I heard of it was in this Matt Parker video where he helped do this with a school-sized group of kids [0]. An "AI player" made of matchboxes, run by schoolkids, is a fantastically fun idea.
Nice. I notice that the author has some other interesting posts. I like this one on the James–Stein estimator [0] and this one [1] on day length variations.
[0] https://joe-antognini.github.io/machine-learning/steins-para... [1] https://joe-antognini.github.io/astronomy/daylight
By the way I think there's a missing factor of rho in the numerator in [1] in the sample transformation section. Should be rho^2
Thank you for the kind words! Yes, I think you're right about the missing factor of rho. And rho^2 is being drawn from a chi-squared distribution, not a chi distribution. (But the mode I stated is correct for a chi-squared distribution --- I must have omitted the squares when typing this up.)
Google produced a podcast introducing ML concepts with the same game a few years ago. https://pair.withgoogle.com/thehardway/
very cool. I love seeing a fully-written-out solution like this, thanks
Beautiful. I'll use this to train an AI on a few of my fave games