Speeding Up Budgeted Stochastic Gradient Descent SVM Training with Precomputed Golden Section Search
This work provides a speed improvement for large-scale SVM training, but it is incremental as it optimizes an existing technique rather than introducing a new paradigm.
The paper tackled the computational bottleneck of iterative budget maintenance in stochastic gradient SVM training by replacing it with a fast lookup method, achieving up to 65% reduction in merging time and 44% reduction in total training time without accuracy loss.
Limiting the model size of a kernel support vector machine to a pre-defined budget is a well-established technique that allows to scale SVM learning and prediction to large-scale data. Its core addition to simple stochastic gradient training is budget maintenance through merging of support vectors. This requires solving an inner optimization problem with an iterative method many times per gradient step. In this paper we replace the iterative procedure with a fast lookup. We manage to reduce the merging time by up to 65% and the total training time by 44% without any loss of accuracy.