Population structure-learned classifier for high-dimension low-sample-size class-imbalanced problem
چکیده
The classification on high-dimension low-sample-size data (HDLSS) is a challenging problem and it common to have class-imbalanced in most application fields. We term this as Imbalanced HDLSS (IHDLSS). Recent theoretical results reveal that the criterion tolerance similarity are crucial HDLSS, which emphasizes maximization of within-class variance premise class separability. Based idea, novel linear binary classifier, termed Population Structure-learned Classifier (PSC), proposed. proposed PSC can obtain better generalization performance IHDLSS by maximizing sum inter-class scatter matrix intra-class separability assigning different intercept values majority minority classes. salient features approach are: (1) It works well IHDLSS; (2) inverse high dimensional be solved low space; (3) self-adaptive determining for each class; (4) Its computational complexity analyzed. A series evaluations conducted one simulated set ten real-world benchmark sets gene analysis. Experimental demonstrate superior state-of-art methods IHDLSS.