cross_validation.ShuffleSplit setting train_size without setting test_size, the sum of train_size and test_size is not equal to 1. Now the sum of them is 0.1(default value of test_size) + train_size.

When setting test_size without setting train_size, the train_size is autocomputed by 1 - test_size, so we hope when setting train_size, the test_size is autocomputed by 1 - train_size as well, not defualt value 0.1.