# Feature learning vs metric learning

What is the fundamental difference between metric learning and feature learning? From feature we can get metric as $M(x,y) = ||f(x) - f(y)||^2$ and vice versa right? (assuming that metric satisfy distance definition)