Large-Scale Multiclass Support Vector Machine Training via Euclidean Projection onto the Simplex

Mathieu Blondel, Akinori Fujino, Naonori Ueda
2014 2014 22nd International Conference on Pattern Recognition  
Dual decomposition methods are the current stateof-the-art for training multiclass formulations of Support Vector Machines (SVMs). At every iteration, dual decomposition methods update a small subset of dual variables by solving a restricted optimization problem. In this paper, we propose an exact and efficient method for solving the restricted problem. In our method, the restricted problem is reduced to the wellknown problem of Euclidean projection onto the positive simplex, which we can solve
more » ... which we can solve exactly in expected O(k) time, where k is the number of classes. We demonstrate that our method empirically achieves state-of-the-art convergence on several large-scale highdimensional datasets.
doi:10.1109/icpr.2014.231 dblp:conf/icpr/BlondelFU14 fatcat:2cm5n5fdrvdm7jkf6rykbxc3zq