website
0.3
Estimating entropy rates with Bayesian confidence intervals
ftp://lyapunov.ucsd.edu/pub/nonlinear/bayesian_information/Kennel_Shlens_et_al_2004.pdf
rating:
not rated
I'm pleased to announce with my co-authors availability of a preprint on our new algorithm to estimate the Shannon entropy rate (bits/symbol) or (bits/sec) of an observed sequence of low-alphabet symbols. It uses the Context-Tree-Weighting universal compression method, but doesnot use the compression ratio directly as an entropy estimator but as a scaffold for a Bayesian estimate. The result is significantly lower bias.
Created: 10 Jul 2004 00:00
by
Mark Nelson
Published: 10 Jul 2004 00:00
Last Updated: 10 Jul 2004 00:00 by
Mark Nelson
Other links to site ucsd.edu:
2 links to this site were found
