1) This is the regression data used in ICML2006 paper.
2) There are 6 datasets {1LJ-16-30-2, 4LJ-80-166-1, 6LJ-160-166-1, 8LJ-16-50-2, 9LJ-80-206-1, 10LJ-160-566-1}
3) For each dataset, the training file is named as "<dataName>.txt", while the testing file is named as "<dataName>-test.txt"
4) The validation set we used in our experiment is also attached: the training file is named as "<dataName>_val_tr.txt", while the testing file is named as "<dataName>_val_test.txt"
5) For the noisy datasets {8LJ-16-50-2, 9LJ-80-206-1, 10LJ-160-566-1} we generated ourselves, the corresponding testsets have not been corrupted by noisy


