VEKTORIŲ KVANTAVIMO METODŲ IR DAUGIAMAČIŲ SKALIŲ JUNGINYS DAUGIAMAČIAMS DUOMENIMS VIZUALIZUOTI
COMBINATION OF VECTOR QUANTIZATION AND MULTIDIMENSIONAL SCALING

Summary/Abstract: In this paper, we present a comparative analysis of a combination of two vector quantization methods (self-organizing map (SOM) and neural gas (NG)), based on neural networks and multidimensional scaling that is used for visualization of codebook vectors obtained by vector quantization methods. The dependence of neuron-winners, quantization and mapping qualities, and preserving of a data structure in the mapping image are investigated. It is established that the quantization errors of NG are smaller than that of the SOM when the number of neurons-winners is approximately equal. It means that the neural gas is more suitable for vector quantization. The data structure is visible in the mapping image even when the number r of neurons-winners of NG is small enough. If the number r of neurons-winners of the SOM is larger, the data structure is visible, as well.

• Details
• Contents