Q-D Relevance

Relevance feedback can be considered a form of “active learning” or “learning with queries”. We Wish to minimize the number of iterations and need to estimate the expected number of iterations remaining. Information theory suggest minimizing the entropy across $P(T=T_i)$.