Domain and Modality Adaptation Using Multi-Kernel Matching

In this paper, we propose a new method for domain and modality adaptation using multi-kernel matching. Our method is based on the representation of the source and target sets with several local kernels centered at a small number of apriori known corresponding samples. We propose to match the local kernels and, in turn, aggregate the local matches and find a mapping between the source and target sets. We showcase the applicability of our method on simulations and real-world data sets that include EEG recordings for mental arithmetic identification and single-cell multi-omics. In these applications, we demonstrate the advantages of our method over recent competing schemes.