After defining the network architecture, the model must be trained. It's now time to define the relationship between the model's output and the real data. To do so, a loss function must be defined.
The loss function is used to assesses the goodness-of-fit of a model.
There are several loss functions, each one expressing a relationship among the network output and the real data, and their form completely influences the quality of the model's prediction.
For a discrete classification problem over 
classes, we can model the defined neural network that accepts a D-dimensional input vector, , and produces an -dimensional vector of ...
