We review the principles of minimum description length and stochastic complexity as used in data compression and statistical modeling. Stochastic complexity is formulated as the solution to optimum universal coding problems extending Shannon's basic source coding theorem. The normalized maximized likelihood, mixture, and predictive codings are each shown to achieve the stochastic complexity to within asymptotically vanishing terms. We assess the performance of the minimum description length criterion both from the vantage point of quality of data compression and accuracy of statistical inference. Context tree modeling, density estimation, and model selection in Gaussian linear regression serve as examples
%0 Journal Article
%1 Barron1998
%A Barron, Andrew R
%A Rissanen, Jorma
%A Yu, Bin
%D 1998
%J IEEE Trans. Inf. Theory
%K information_theory mdl model schema phd schemdesc
%N 6
%P 2743--2760
%T The Minimum Description Length Principle in Coding and Modeling
%U http://dblp.uni-trier.de/rec/bibtex/journals/tit/BarronRY98
%V 44
%X We review the principles of minimum description length and stochastic complexity as used in data compression and statistical modeling. Stochastic complexity is formulated as the solution to optimum universal coding problems extending Shannon's basic source coding theorem. The normalized maximized likelihood, mixture, and predictive codings are each shown to achieve the stochastic complexity to within asymptotically vanishing terms. We assess the performance of the minimum description length criterion both from the vantage point of quality of data compression and accuracy of statistical inference. Context tree modeling, density estimation, and model selection in Gaussian linear regression serve as examples
@article{Barron1998,
abstract = {We review the principles of minimum description length and stochastic complexity as used in data compression and statistical modeling. Stochastic complexity is formulated as the solution to optimum universal coding problems extending Shannon's basic source coding theorem. The normalized maximized likelihood, mixture, and predictive codings are each shown to achieve the stochastic complexity to within asymptotically vanishing terms. We assess the performance of the minimum description length criterion both from the vantage point of quality of data compression and accuracy of statistical inference. Context tree modeling, density estimation, and model selection in Gaussian linear regression serve as examples},
added-at = {2013-12-17T09:48:27.000+0100},
author = {Barron, Andrew R and Rissanen, Jorma and Yu, Bin},
biburl = {https://www.bibsonomy.org/bibtex/2a41594da2b9b16cddb90fc163acd6df7/jullybobble},
file = {:Users/julien.gaugaz/Dropbox/Papers/Mendeley Desktop/1998/Barron, Rissanen, Yu - 1998 - The Minimum Description Length Principle in Coding and Modeling.pdf:pdf},
interhash = {f8eb1976d2b5762d746808c0fb4e83ae},
intrahash = {a41594da2b9b16cddb90fc163acd6df7},
journal = {IEEE Trans. Inf. Theory},
keywords = {information_theory mdl model schema phd schemdesc},
number = 6,
pages = {2743--2760},
timestamp = {2014-07-27T15:43:19.000+0200},
title = {{The Minimum Description Length Principle in Coding and Modeling}},
url = {http://dblp.uni-trier.de/rec/bibtex/journals/tit/BarronRY98},
volume = 44,
year = 1998
}