Fisher's Discriminant is a linear classifier used in statistics and machine learning to find a linear combination of features that separates two or more classes of objects or events. It maximizes the ratio of the variance between classes to the variance within classes, thereby achieving the best class separability in a reduced-dimensional space.