A team is comparing two approaches for classifying scanned invoices. One uses a classical model where a data scientist defines the inputs, such as the number of table rows and the position of the total. The other uses a deep neural network trained on the raw images. Which statement best describes the difference?

• AThe deep network will always be more accurate because it has more parameters.
• BThe classical model is a form of deep learning applied to structured data.
• COnly the classical approach requires labelled training examples.
• DThe deep network learns its own features from the raw input, whereas the classical model depends on features a person defined.