Analysis of Robust Soft Learning Vector Quantization and an application to Facial Expression Recognition

Learning Vector Quantization (LVQ) [1] is a popular method for multiclass classification. Several variants of LVQ have been developed recently, of which Robust Soft Learning Vector Quantization (RSLVQ) [2] is a promising one. Although LVQ methods have an intuitive design with clear updating rules, their dynamics are not yet well understood. In simulations within a controlled environment RSLVQ performed very close to optimal. This controlled environment enabled us to perform a mathematical analysis as a first step in obtaining a better theoretical understanding of the learning dynamics. This extended abstract provides the outline of our theoretical analysis and its results. Moreover, we will focus on the practical application of RSLVQ to a real world data set containing extracted features from facial expression data.