Dynamics of Multiple Prototype LVQ

Learning Vector Quantization (LVQ) is a family of prototype-based multi-class classification algorithms. LVQ has been used in various fields, including medical image or data analysis [1]. A mathematical description of LVQ is useful to evaluate performance of different algorithms and develop novel and efficient schemes. The dynamics of VQ, LVQ1, LFM, RSLVQ and LVQ2.1 have been investigated for 2 prototypes and 2 classes [4]. Multi-prototype, multi-class problems produce a more general and realistic study on the behavior of LVQ algorithms.