What:

It’s a way of representing the feed-forward pass of a neural network where operations (add, multiply, etc.), are nodes and operands are incoming edges.

Example:

The below is a 1-dimensional neuron with MSE loss.