Multi-Merge Budget Maintenance for Stochastic Gradient Descent SVM Training

06/26/2018

∙

by   Sahar Qaadan, et al.

∙

0

∙

share

Budgeted Stochastic Gradient Descent (BSGD) is a state-of-the-art technique for training large-scale kernelized support vector machines. The budget constraint is maintained incrementally by merging two points whenever the pre-defined budget is exceeded. The process of finding suitable merge partners is costly; it can account for up to 45 paper we investigate computationally more efficient schemes that merge more than two points at once. We obtain significant speed-ups without sacrificing accuracy.