How to visualize a classifier?

Aug, 2012

We propose a general framework how to nonlinearly visualize a given classifier and training set provided the classifier, besides the crisp classification result, offers smooth values indicating the distance from the nearest class boundary. We discuss the requirements which this framework poses on a nonlinear dimensionality reduction method used for the visualization

@inproceedings{Schulz_NC_2_2012,
author = {Alexander Schulz and Andrej Gisbrecht and Kerstin Bunte and Barabara Hammer},
title = {How to visualize a classifier?},
booktitle = {New Challenges in Neural Computation (NC$^{2}$)},
series = {Workshop of the GI-Fachgruppe Neuronale Netze and the German Neural Networks Society in connection to DAGM 2012},
pages = {73--83},
month = {"Aug."},
editor = {Barbara Hammer and Thomas Villmann},
address = {Graz, Austria},
publisher = {LNCS},
year = {2012},
abstract = {We propose a general framework how to nonlinearly visualize a given classifier and training set provided the classifier,  besides the crisp classification result, offers smooth values indicating the distance from the nearest class boundary.  We discuss the requirements which this framework poses on a nonlinear dimensionality reduction method used for the visualization},
}