Comment by wmasson

It's just something I wrote myself. Fairly adhoc - it just checks through the latex notes for theorems and keeps them in a list with a current score. The picking is done using softmax selection which picks randomly with a bias towards very low scores. When I complete a theorem I enter how well I did at recalling it and then it updates the score using temporal difference learning. I'm going to clean up the code and put something like it on github sometime.