Local metric adaptation for soft nearest prototype classification to classify proteomic data

We propose a new method for the construction of nearest prototype classifiers which is based on a Gaussian mixture approach interpreted as an annealed version of Learning Vector Quantization. Thereby we allow the adaptation of the underling metric which is useful in proteomic research. The algorithm performs a gradient descent on a cost function adapted from soft nearest prototype classification. We investigate the properties of the algorithm and assess its performance on two clinical cancer data sets. Results show that the algorithm performs reliable with respect to alternative state of the art classifiers.