Learns a linear SVM trained with stochastic gradient descent, based on the implementation in scikit-learn, for classification of binary-labeled data. Can be used for training or testing, depending on the arguments passed in:
For training: takes in a data collection file path to write a trained model to and two other data collection paths to read data and labels in from, along with hyperparameter settings (in order: regularization coefficient, initial learning rate, learning rate decay parameter, and number of iterations of stochastic gradient descent). Writes the trained model to the file at the specified data collection path.
For testing: takes in one path to a file in a data collection that contains a trained model to use, another file path to a test data set, and a third file path to which to write predictions.