The following are the aguments for running the experiments on the auto-encoder example of functional regularization:

repr_epochs : Number of Training Epochs for the Auxiliary Task

pred_epochs : Number of Training Epochs for the Prediction Task

repr_size : Number of unlabeled data points for the Auxiliary Task

pred_size : Number of labeled data points for the Prediction Task

num_trials : Number of independent runs to average scores over

reg_w1 : Weight on the orthonormal regularization on W for h

reg_w2 : Weight on the orthonormal regularization on V for g

repr_lr : Learning rate for the Auxiliary Task

pred_lr : Learning rate for the Prediction Task

repr_static : Make the representation h learned from the Auxiliary task static for the Prediction Task

r : Dimension r

d : Dimension of the data d

For running the code, see example in run.sh