zbMATH — the first resource for mathematics

Stochastic dual coordinate ascent methods for regularized loss minimization. (English) Zbl 1307.68073
Summary: Stochastic Gradient Descent (SGD) has become popular for solving large scale supervised machine learning optimization problems such as SVM, due to their strong theoretical guarantees. While the closely related Dual Coordinate Ascent (DCA) method has been implemented in various software packages, it has so far lacked good convergence analysis. This paper presents a new analysis of Stochastic Dual Coordinate Ascent (SDCA) showing that this class of methods enjoy strong theoretical guarantees that are comparable or better than SGD. This analysis justifies the effectiveness of SDCA for practical applications.

68T05 Learning and adaptive systems in artificial intelligence
68Q25 Analysis of algorithms and problem complexity
Full Text: Link arXiv