Yes the Answer is a) True
For an over fit data set the cross validation error will be much bigger than the training error because than the validation error in the system the training error is much lower as the because our model is to get over fit.But if the dropout regularization layer is being used in the network then the validation layer becomes smaller than the training error.