Back propagation refers to a type of neural network model in which the system uses information about the appropriateness of its output to adjust the weightings of the connections among intervening units.