Belief Propagation: Accurate Marginals or Accurate Partition Function - Where is the Difference?

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

Abstract

We analyze belief propagation on patch potential models -- these are attractive models with varying local potentials -- obtain all of the possibly many fixed points, and gather novel insights into belief propagation's properties. In particular, we observe and theoretically explain several regions in the parameter space that behave fundamentally different. We specify and elaborate on one specific region that, despite the existence of multiple fixed points, is relatively well behaved and provides insights into the relationship between the accuracy of the marginals and the partition function. We demonstrate the inexistence of a principle relationship between both quantities and provide sufficient conditions for a fixed point to be optimal with respect to approximating both the marginals and the partition function.
Original languageEnglish
Title of host publicationUncertainty in Artificial Intelligence
Publication statusPublished - 23 Jul 2019
Event2019 Conference on Uncertainty in Artificial Intelligene - Tel Aviv, Israel
Duration: 22 Jul 201925 Jul 2019

Conference

Conference2019 Conference on Uncertainty in Artificial Intelligene
Abbreviated titleUAI 2019
CountryIsrael
CityTel Aviv
Period22/07/1925/07/19

Fingerprint

partitions
propagation

Cite this

Belief Propagation: Accurate Marginals or Accurate Partition Function - Where is the Difference? / Knoll, Christian; Pernkopf, Franz.

Uncertainty in Artificial Intelligence. 2019.

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

Knoll, C & Pernkopf, F 2019, Belief Propagation: Accurate Marginals or Accurate Partition Function - Where is the Difference? in Uncertainty in Artificial Intelligence. 2019 Conference on Uncertainty in Artificial Intelligene, Tel Aviv, Israel, 22/07/19.
@inproceedings{eb4eea44985343349f1dab10f3b877d0,
title = "Belief Propagation: Accurate Marginals or Accurate Partition Function - Where is the Difference?",
abstract = "We analyze belief propagation on patch potential models -- these are attractive models with varying local potentials -- obtain all of the possibly many fixed points, and gather novel insights into belief propagation's properties. In particular, we observe and theoretically explain several regions in the parameter space that behave fundamentally different. We specify and elaborate on one specific region that, despite the existence of multiple fixed points, is relatively well behaved and provides insights into the relationship between the accuracy of the marginals and the partition function. We demonstrate the inexistence of a principle relationship between both quantities and provide sufficient conditions for a fixed point to be optimal with respect to approximating both the marginals and the partition function.",
author = "Christian Knoll and Franz Pernkopf",
year = "2019",
month = "7",
day = "23",
language = "English",
booktitle = "Uncertainty in Artificial Intelligence",

}

TY - GEN

T1 - Belief Propagation: Accurate Marginals or Accurate Partition Function - Where is the Difference?

AU - Knoll, Christian

AU - Pernkopf, Franz

PY - 2019/7/23

Y1 - 2019/7/23

N2 - We analyze belief propagation on patch potential models -- these are attractive models with varying local potentials -- obtain all of the possibly many fixed points, and gather novel insights into belief propagation's properties. In particular, we observe and theoretically explain several regions in the parameter space that behave fundamentally different. We specify and elaborate on one specific region that, despite the existence of multiple fixed points, is relatively well behaved and provides insights into the relationship between the accuracy of the marginals and the partition function. We demonstrate the inexistence of a principle relationship between both quantities and provide sufficient conditions for a fixed point to be optimal with respect to approximating both the marginals and the partition function.

AB - We analyze belief propagation on patch potential models -- these are attractive models with varying local potentials -- obtain all of the possibly many fixed points, and gather novel insights into belief propagation's properties. In particular, we observe and theoretically explain several regions in the parameter space that behave fundamentally different. We specify and elaborate on one specific region that, despite the existence of multiple fixed points, is relatively well behaved and provides insights into the relationship between the accuracy of the marginals and the partition function. We demonstrate the inexistence of a principle relationship between both quantities and provide sufficient conditions for a fixed point to be optimal with respect to approximating both the marginals and the partition function.

M3 - Conference contribution

BT - Uncertainty in Artificial Intelligence

ER -