Geometric Metric Learning for Multi-Output Learning

Due to its wide applications, multi-output learning that predicts multiple output values for a single input at the same time is becoming more and attractive. As one of most popular frameworks dealing with learning, performance k-nearest neighbor (kNN) algorithm mainly depends on metric used compute distance between different instances. In this paper, we propose novel cost-weighted geometric mean method learning. Specifically, learns which can make embedding correct be smaller than outputs nearest neighbors. The learned discover dependencies move instances far away in space. addition, our objective function has closed solution, thus calculation speed very fast. Compared state-of-the-art methods, it easier explain also faster speed. Experiments conducted two tasks (i.e., multi-label classification multi-objective regression) have confirmed provides better results methods.