Fraunhofer-Gesellschaft

Publica

Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Archetypal analysis as an autoencoder

 
: Bauckhage, C.; Kersting, K.; Hoppe, F.; Thurau, C.

:
Volltext (PDF; )

Hammer, B. ; Gesellschaft für Informatik -GI-, Fachgruppe Neuronale Netze:
Workshop New Challenges in Neural Computation, NC2 2015 : In connection to GCPR 2015, October 07 - 10, 2015, Aachen, Germany
Aachen, 2015 (Machine Learning Reports 03/2015)
S.8-15
Workshop "New Challenges in Neural Computation" (NC2) <6, 2015, Aachen>
German Conference on Pattern Recognition (GCPR) <37, 2015, Aachen>
Englisch
Konferenzbeitrag, Elektronische Publikation
Fraunhofer IAIS ()

Abstract
We present an efficient approach to archetypal analysis where we use sub-gradient algorithms for optimization over the simplex to determine archetypes and reconstruction coefficients. Runtime evaluations reveal our approach to be notably more efficient than previous techniques. As an practical application, we consider archetypal analysis for autoencoding.

: http://publica.fraunhofer.de/dokumente/N-415147.html