Регуляризационная константа

Supervised topic models with a logistic likelihood have two issues that potential- ly limit their practical use: 1) response variables are usually over-weighted by document word counts; and 2) existing …...