10.07.2015 Views

Funcionamiento, interfaz y formato de los datos en MaxEnt - Gbif.es

Funcionamiento, interfaz y formato de los datos en MaxEnt - Gbif.es

Funcionamiento, interfaz y formato de los datos en MaxEnt - Gbif.es

SHOW MORE
SHOW LESS
  • No tags were found...

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

AbstractFinite-valued input two-dim<strong>en</strong>sional (2-D) Gaussian channels with memoryrepr<strong>es</strong><strong>en</strong>t an important class of systems, which appears ext<strong>en</strong>sively incommunications and storage. In spite of their wi<strong>de</strong>spread use, the workingsof 2-D channels are still very much unknown. In this work we tryto explore their properti<strong>es</strong> from the perspective of <strong>es</strong>timation theory andinformation theory.At the heart of our approach is a mapping of a 2-D channel to anundirected graphical mo<strong>de</strong>l, and inferring its a-posteriori probabiliti<strong>es</strong>using g<strong>en</strong>eralized belief propagation (GBP). The <strong>de</strong>rived probabiliti<strong>es</strong>are shown to be practically accurate, thus <strong>en</strong>abling optimal maximuma-posteriori (MAP) <strong>es</strong>timation of the transmitted symbols. Also, theShannon-theoretic information rat<strong>es</strong> are <strong>de</strong>duced either via the vector-wiseShannon-McMillan-Breiman theorem, or via the rec<strong>en</strong>tly <strong>de</strong>rived symbolwiseGuo-Shamai-Verdú theorem. Our approach is also <strong>de</strong>scribed from theperspective of statistical mechanics, as the graphical mo<strong>de</strong>l and infer<strong>en</strong>cealgorithm have their analogu<strong>es</strong> in physics.Our experim<strong>en</strong>tal study, based on common channel settings tak<strong>en</strong> fromcellular networks and magnetic recording <strong>de</strong>vic<strong>es</strong>, <strong>de</strong>monstrat<strong>es</strong> that un<strong>de</strong>rnon-trivial memory conditions, the performance of this fully tractableGBP <strong>es</strong>timator is almost i<strong>de</strong>ntical to the performance of the optimal MAP<strong>es</strong>timation. It also <strong>en</strong>abl<strong>es</strong> a practically accurate simulation-based <strong>es</strong>timateof the information rate. Rationalization of this excell<strong>en</strong>t performanceof GBP in the 2-D Gaussian channel setting is addr<strong>es</strong>sed.In<strong>de</strong>x Terms: Two-dim<strong>en</strong>sional channels, intersymbol interfer<strong>en</strong>ce, maximuma-posteriori <strong>es</strong>timation, information rate, g<strong>en</strong>eralized belief propagation, clustervariation method, Shannon-McMillan-Breiman theorem, Guo-Shamai-Verdútheorem, magnetic recording channels, multiple-acc<strong>es</strong>s channels.2

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!