A Fused Elastic Net Logistic Regression Model for Multi-Task Binary Classification

摘要

Multi-task learning has shown to significantly enhance the performance ofmultiple related learning tasks in a variety of situations. We present thefused logistic regression, a sparse multi-task learning approach for binaryclassification. Specifically, we introduce sparsity inducing penalties overparameter differences of related logistic regression models to encodesimilarity across related tasks. The resulting joint learning task is cast intoa form that lends itself to be efficiently optimized with a recursive variantof the alternating direction method of multipliers. We show results onsynthetic data and describe the regime of settings where our multi-taskapproach achieves significant improvements over the single task learningapproach and discuss the implications on applying the fused logistic regressionin different real world settings.