Automatic differentiation for neighbourhood component analysis
A tutorial on how to apply automatic differentiation to efficiently compute the gradients of a function. We consider the case of a metric learning algorithm—neighbourhood component analysis.