Desktop Survival Guide
by Graham Williams
An issue with support vector machines is that parameter tuning is not altogether easy for users new to them. One computationally expensive approach is to build multiple models with different parameters and choose the one with lowest expected error. This can lead to suboptimal results though, unless quite extensive search is performed. Research has explored using previous performance of different parameter settings to predict their relative performance on new tasks [#!soares.etal:2004:meta_learn_svm!#].