Optimally Spaced Repetition Learning

One of the ideas I've been toying with in my head is writing a chess training software (see CT-ART 3) that not only utilizes the computer as a training tool for chess tactics, but also, that incorporates the idea of optimally spaced repetition learning (aka SuperMemo algorithm).

By defining a set of of question/answer type tactics problems along with an algorithm that models human memory, a more effective and optimal approach to learning the problem sets and retaining the content can be acheived.

I would like to use the SuperMemo-2 algorithm along with a proven and solved set of tactics problems and I would implement it in Java using Chesspresso.

This shall prove to be in interesting experiment in learning ;-).