# CODE TO COMPARE RISK BOUNDS AND POST-TRAINING PERFORMANCE FOR REGRESSION
# USING META-ALGO vs SGD+TEST-SET

# HOW TO: to run the algorithms, simply run 'runexp.py'

# NOTES: 
#        1. 'runexp.py' produces the results from one dataset and one
#        pretraining fraction. To plot the results as in Figure 4
#        the code need be executed multiple times over all datasets 
#        and pretrain fractions.
#        To do so, simply change DATASET_ID and FRACTION_PRETRAIN in 'runexp.py'
#
#        2. 'create_and_save_dataset.py' was used to create the datasets 
#        dataset_test_sinc(2.5x)_sigma_0.05.npy
#        dataset_train_sinc(2.5x)_sigma_0.05.npy
#        This is included for completeness only and need not be run