Model Reuse With Reduced Kernel Mean Embedding Specification

Given a publicly available pool of machine learning models constructed for various tasks, when user plans to build model her own application, is it possible upon in the such that previous efforts on these existing can be reused rather than starting from scratch? Here, grand challenge how find are helpful current without accessing raw training data pool. In this paper, we present two-phase framework. upload phase, uploading into pool, construct reduced kernel mean embedding (RKME) as specification model. Then deployment relatedness task and pre-trained will measured based value RKME specification. Theoretical results extensive experiments validate effectiveness our approach.