Accelerated proximal stochastic dual coordinate ascent for regularized loss minimization (Q5962715)

A minimization problem related to the machine learning is considered. The objective function is a regularized loss function obtained by adding a regularizer to the convex loss function. A new version of the stochastic dual coordinate ascent method is proposed and its high convergence rate is proven. This result enables improvement of the key machine learning algorithms including support vector machines (SVM), ridge regression, Lasso, and multiclass SVM. The experimental results are included which corroborate the theoretical findings.