paper available on LBG-U
Bernd Fritzke
fritzke at neuroinformatik.ruhr-uni-bochum.de
Thu Jan 9 12:49:47 EST 1997
ftp://ftp.neuroinformatik.ruhr-uni-bochum.de/pub/manuscripts/IRINI/irini97-01/irini97-01.ps.gz
The following TR/preprint is available via ftp (93 KB, 10 pages):
The LBG-U method for vector quantization -
an improvement over LBG inspired from neural networks
Bernd Fritzke
Systembiophysik
Institut f"ur Neuroinformatik
Ruhr-Universit"at Bochum * Germany
(to appear in: Neural Processing Letters, 1997, Vol. 5, No. 1)
Keywords: codebook construction, data compression, growing neural
networks, LBG, vector quantization
Abstract:
A new vector quantization method -- denoted LBG-U -- is presented which is
closely related to a particular class of neural network models (growing
self-organizing networks). LBG-U consists mainly of repeated runs of the
well-known LBG algorithm. Each time LBG has converged, however, a novel
measure of utility is assigned to each codebook vector. Thereafter, the
vector with minimum utility is moved to a new location, LBG is run on the
resulting modified codebook until convergence, another vector is moved, and
so on. Since a strictly monotonous improvement of the LBG-generated
codebooks is enforced, it can be proved that LBG-U terminates in a finite
number of steps. Experiments with artificial data demonstrate significant
improvements in terms of RMSE over LBG combined with only modestly higher
computational costs.
Comments are welcome,
Bernd Fritzke
PS: Sorry for the long and obviously redundant URL. In some cases our
TRs are provided as LaTeX source with all figures in separate
files. Therefore, each TR has its own directory. Its a German system 8v).
--
Bernd Fritzke * Institut f"ur Neuroinformatik Tel. +49-234 7007845
Ruhr-Universit"at Bochum * Germany FAX. +49-234 7094210
WWW: http://www.neuroinformatik.ruhr-uni-bochum.de/ini/PEOPLE/fritzke/top.html
More information about the Connectionists
mailing list