Get the latest tech news

Solving Redactle with Decision Trees

me in which you are presented with a Wikipedia page in which most words are redacted, and you need to find the article's title. You can submit as many words as you want, and if the word you submitted is anywhere in the article, it "unredacts" all the occurrences of it, so you can progressively comprehend what the article is about by trial and error.

And the algorithm would automatically find the best flowchart (according to some definition of "best") to allow you to classify any animals (depending on its features) you present to it into one of the target "class". Remember, it doesn't necessarily build a balanced tree as it can shortcut to small homogeneous classes, so the number of leaves is expected to be lesser or equal to 2**max_depth. Note that there is an element of randomness in the generation of the tree (which I didn't take the time to investigate), so those specific values are not perfectly reproducible.

Get the Android app