Consider a decision tree for the handwritten digit recognition problem, we get low training error, but large test error. The problem with our decision tree has happened because

Lote of training data

Decision tree is under fitting

None of the above

Decision tree is very deep