AdaptiveMulti-stageDensityRatioEstimationfor LearningLatentSpaceEnergy-basedModel

Toeffectively tackle this issue and learn more expressiveprior models, wedevelop theadaptivemulti-stage density ratio estimation which breaks the estimation into multiple stages and learn different stages ofdensity ratiosequentially andadaptively. Thelatent priormodel canbe gradually learned using ratio estimated in previous stage so that the final latent spaceEBMpriorcanbenaturally formed byproduct ofratiosindifferentstages. The proposed method enables informativeand much sharper prior than existing baselines, and can be trained efficiently.