Easy-But-Effective Domain Sub-Similarity Learning for Transfer Regression

Transfer covariance function, which can model domain similarity and adaptively control the knowledge transfer across domains, is widely used in learning. In this paper, we concentrate on Gaussian process ( GP ) models using a function for regression problems black-box learning scenario. Precisely, investigate family of rather general functions, ${T}_{*}$ , that heterogeneous sub-similarities domains through multiple kernel A necessary sufficient condition to obtain valid s notation="LaTeX">$GP_{T_{*}}$ any data given. This becomes specially handy practical applications as (i) it enables semantic interpretations (ii) readily be particular, propose computationally inexpensive rule explicitly capture different domains. We two instantiations one with set predefined constant base kernels learnable parametric kernels. Extensive experiments 36 synthetic tasks 12 real-world demonstrate effectiveness sub-similarity performance.