Abstract
Given multi-category point sets from different place-types, our goal is todevelop a spatially-lucid classifier that can distinguish between two classesbased on the arrangements of their points. This problem is important for manyapplications, such as oncology, for analyzing immune-tumor relationships anddesigning new immunotherapies. It is challenging due to spatial variability andinterpretability needs. Previously proposed techniques require dense trainingdata or have limited ability to handle significant spatial variability within asingle place-type. Most importantly, these deep neural network (DNN) approachesare not designed to work in non-Euclidean space, particularly point sets.Existing non-Euclidean DNN methods are limited to one-size-fits-all approaches.We explore a spatial ensemble framework that explicitly uses different trainingstrategies, including weighted-distance learning rate and spatial domainadaptation, on various place-types for spatially-lucid classification.Experimental results on real-world datasets (e.g., MxIF oncology data) showthat the proposed framework provides higher prediction accuracy than baselinemethods.