what is backward propagation in neural network

Recipe Objective - What is backward propagation?

Backward propagation is supervised algorithm used for training neural networks(Multi-layer perceptrons). It uses gradient descent or delta rule to look for the minimum error functions in the weight space and selects weights that minimize the error solution. Backward propagation is also known as Backpropagation. It is a method of fine-tuning the weights of neural network on the basis of error rate obtained in previous epoch. Error rates are reduced using proper tuning of weights and it increases generalization of model.

Explanation of Backward propagation.

Backward propagation is a fast and simple program which contains inputs and no parameters for tuning.

Backward propagation is a standard and flexible method as prior network knowledge is not necessary and it generally works good.

Backward propagation doesn't require function features to be learned.