Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Calibrated Simplex Mapping Classification

Published on arXiv
: Heese, Raoul; Walczak, Michał; Bortz, Michael; Schmid, Jochen

Fulltext ()

2021, arXiv:2103.02926, 24 pp.
Paper, Electronic Publication
Fraunhofer ITWM ()

We propose a novel supervised multi-class/single-label classifier that maps training data onto a linearly separable latent space with a simplex-like geometry. This approach allows us to transform the classification problem into a well-defined regression problem. For its solution we can choose suitable distance metrics in feature space and regression models predicting latent space coordinates. A benchmark on various artificial and real-world data sets is used to demonstrate the calibration qualities and prediction performance of our classifier.