Support model parameters

[Leengit]

Allow the user to override our default parameters for machine learning models, including to select number of epochs, optimizer, batch size, training weights, loss function, and whether final layer of model is softmax or log_softmax, In the case of Bayesian machine learning models, user should also be able to override num_train_samples, num_validation_samples, num_predict_samples.