Deep learning methods using deep neural networks can effectively solve problems while requiring significantly less training data. Additionally, the creation of deep architecture allows for the creation of an admissible set of functions based on the available training data.