Why is it important to have separate training/validation and testing tests?

A. Some parameters that are chosen using the validation data may not make sense on the training data.

B. If one tests on the training data, it is impossible to know whether the algorithm will generalize to new data.

C. The testing set can be very large, so it makes sense to have smaller training/validation sets that can be processed quickly as one figures out model parameters.

D. None of the above