Efficient Implementation of Optimal Linear Discriminant Analysis

Abstract

Here we aim at optimizing the implementation of Fisher’s Linear Discriminant Analysis (FLDA) - based classification for the small sample size problem. Among methods for this problem it considers the one closest, in some sense, to the classical regular approach and improves its implementation with regards to computational and storage costs and numerical stability.

Keywords

• efficient
• implementation
• optimal
• linear
• discriminant
• analysis