Add an overload to TrainingData to accept only 1 argument, the input vector

.use cases GVF and DTW that do not require the vector of outputs or labels

.refactor the clients of the function that are using empty output vectors