Infer.NET user guide : Learners : Bayes Point Machine classifiers : Command-line runners
A Bayes Point Machine is trained using the
Train module, both in binary and multi-class classification. The
Train module reads a training set and returns a serialized trained classifier, which can then be used to make predictions or train incrementally.
Train module has the following command-line arguments:
training-set: The file with training data containing ground truth labels and features in the format described earlier.
model: The file to which the trained Bayes Point Machine classifier will be saved.
iterations: The number of training algorithm iterations (defaults to 30).
batches: The number of batches into which the training data is split (defaults to 1).
compute-evidence: If specified, the Bayes Point Machine classifier will compute model evidence on the training data (defaults to false).
For more information about the command-line arguments, see Settings. A more detailed explanation of training is available here.
Learner Classifier BinaryBayesPointMachine Train --training-set training-set.dat --model trained-binary-bpm.bin --iterations 25 --batches 2 --compute-evidence Learner Classifier MulticlassBayesPointMachine Train --training-set training-set.dat --model trained-multiclass-bpm.bin --iterations 25 --batches 2 --compute-evidence