Backpropagation is a rather basic concept of machine learning. I want to write a concise review here.

Classes review version

Andrew NG

$a^{(i)}$ is output, $z^{(i)}\ \text{or}\ x$ is input, $\theta ^{( i )}$ is weight, $g(x)$ is activation function.

Compute the delta terms for the earlier layer.

Training a neural network steps

Reference

[1] Backpropagation

[2] Andrew NG, Machine learning