I have been working on a project with LibSVM and have noticed there is an option to train the SVM model with "shrinking heuristics" which are used to speed up the classifier training.
After doing some googling, I couldn't find anything substantial explaining what shrinking heuristics is. My understanding is that shrinkage is the changing of parameter estimates to be closer to each other, and heuristics are rules for decision making. So I would think that shrinking heuristics in the LibSVM context have something to do with reducing the distance between the SV's in order to classify the data. With that said, I don't understand how such a tool can make training faster, and question whether this would affect the accuracy of the model.
Could someone provide some insight as to what shrinking heuristics is both generally and in the context of LibSVM. Any links to further explain this would be greatly appreciated too.
Best Answer
This shrinking method is explained in section 5.1 of the official documentation
[1]. You can find the pdf here.[1]C.-C. Chang and C.-J. Lin. LIBSVM : a library for support vector machines. ACM Transactions on Intelligent Systems and Technology, 2:27:1--27:27, 2011