The most common learning tasks for neural networks are approximation, classification, and forecasting.
The data set contains the information available for creating the model.
The neural network represents the approximation or classification model.
The training strategy fits the neural network with the data set.
Model selection algorithms look for the neural network architecture with the best generalization capabilities.
The testing analysis compares the outputs from the neural network against the targets in an independent set.
Model deployment is used to apply a model to predict new data.