The SparseSVMPredictor function takes the model generated by the function SparseSVMTrainer and a set of test samples (in sparse format) and outputs a prediction for each sample.
This function can be used with real-time applications. See AMLGenerator.