# Contrastive Methods

Contrastive methods learn representations by contrasting positive and negative examples. This is one of methods used in Self-supervised Learning.

The goal is to learn an encoder \(f\) such that:

\begin{equation} \text{score}(f(x), f(x^{+})) » \text{score}(f(x), f(x^{-})) \end{equation}

where \(x^{+}\) is a positive example, and \(x^{-}\) is a negative example.