Skip to main navigation Skip to search Skip to main content

From data to constraints

  • S. Mukhopadhyay
  • , E. Parzen
  • , S. N. Lahiri

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Jaynes' Maximum Entropy (MaxEnt) inference starts with the assumption that we have a set of known constraints over the distribution. In statistical physics, we have a good intuition about the conserved macroscopic variables. It should not be surprising that in a real world applications, we have no idea about which coordinates to use for specifying the state of the system. In other words, we only observe empirical data and we have to take a decision on the constraints from the data. In an effort to circumvent this limitation, we propose a nonparametric quantile based method to extract relevant and significant facts (sufficient statistics) for the maximum entropy exponential model.

Original languageEnglish
Title of host publicationBayesian Inference and Maximum Entropy Methods in Science and Engineering - 31st International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2011
Pages32-39
Number of pages8
DOIs
StatePublished - 2012
Event31st International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2011 - Waterloo, ON, Canada
Duration: Jul 9 2011Jul 16 2011

Publication series

NameAIP Conference Proceedings
Volume1443
ISSN (Print)0094-243X
ISSN (Electronic)1551-7616

Conference

Conference31st International Workshop on Bayesian Inference and Maximum Entropy Methods in Science and Engineering, MaxEnt 2011
Country/TerritoryCanada
CityWaterloo, ON
Period07/9/1107/16/11

Keywords

  • exponential model
  • Maximum entropy
  • mid-rank transformations
  • nonparametric Entropy estimation
  • quantile function

Fingerprint

Dive into the research topics of 'From data to constraints'. Together they form a unique fingerprint.

Cite this