Supervised Neural Gas for Learning Vector Quantization
T. Villmann, Barbara Hammer · 2002
In this contribution we combine approaches the generalized leraning vector quantization (GLVQ) with the neighborhood oftentented learning in the neural gas network (NG). In this way we obtain a supervised version of the NG what we call supervised NG (SNG). We show that the SNG is more robust than the GLVQ because the neighborhood learning avoids numerically instabilities as it may occur for complicate classification tasks like in the case of multimodal data.