GA-MTL:A RandomMethodofMulti-Task Learn'ing
T. Y. LiuG, Guojun Wu, E. C. · 2006
Multi-task learning techniques canemploythe removed redundant information toimprove prediction accu- racy.Whichfeatures toaddtothetarget and/or theinput during multi-task learning isstill anopenissue. Theprevious study usedheuristic search methods. Inthis paper, arandom methodofgenetic algorithm basedmulti-task learning (GA- MTL)isproposed toautomatically determine thefeatures for theinput and/or thetarget. Experimental results ondatasets fromthereal worldshowthatGA-MTLiseasytouseandob- tains better performance thanheuristic methods. Keywords-Multi-Task Learning; Genetic Algorithm; Feature