背景介绍TransferLearning-JDL联合试验室.PPT

* * * * * * * Then given a new sample w={ω1, ω2,…., ωn},we can calculate p(w|x=l),if x corresponding label info * * * * * * * * * * * * * * * * * * * * * * * * * * * * Method to Transfer Importance of informative priors for zero-shot learning Category-Topic Model Have to see some training samples of target category Otherwise, can only give symmetric Dirichlet prior Cannot be used in zero-shot learning Category-Topic Model with informative priors give Dirichlet prior Can be used in zeros-shot one-shot learning Outline 文章介绍 背景介绍 Transfer Learning Author-Topic Model Attribute-Based Transf