Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows

Biwei HUANG, Kun ZHANG, Jiji ZHANG, Ruben SANCHEZ-ROMERO, Clark GLYMOUR, Bernhard SCHÖKOPF

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

3 Citations (Scopus)

Abstract

We address two important issues in causal discovery from nonstationary or heterogeneous data, where parameters associated with a causal structure may change over time or across data sets. First, we investigate how to efficiently estimate the "driving force" of the nonstationarity of a causal mechanism. That is, given a causal mechanism that varies over time or across data sets and whose qualitative structure is known, we aim to extract from data a low-dimensional and interpretable representation of the main components of the changes. For this purpose we develop a novel kernel embedding of nonstationary conditional distributions that does not rely on sliding windows. Second, the embedding also leads to a measure of dependence between the changes of causal modules that can be used to determine the directions of many causal arrows. We demonstrate the power of our methods with experiments on both synthetic and real data.
Original languageEnglish
Title of host publication2017 IEEE International Conference on Data Mining (ICDM)
PublisherIEEE
Pages913-918
Publication statusPublished - 2017
EventIEEE International Conference on Data Mining - United States, New Orleans, United States
Duration: 18 Nov 201721 Nov 2017

Publication series

Name
ISSN (Electronic)2374-8486

Conference

ConferenceIEEE International Conference on Data Mining
CountryUnited States
CityNew Orleans
Period18/11/1721/11/17
OtherIEEE computer society

Fingerprint

Experiments

Cite this

HUANG, B., ZHANG, K., ZHANG, J., SANCHEZ-ROMERO, R., GLYMOUR, C., & SCHÖKOPF, B. (2017). Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows. In 2017 IEEE International Conference on Data Mining (ICDM) (pp. 913-918). IEEE.
HUANG, Biwei ; ZHANG, Kun ; ZHANG, Jiji ; SANCHEZ-ROMERO, Ruben ; GLYMOUR, Clark ; SCHÖKOPF, Bernhard. / Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows. 2017 IEEE International Conference on Data Mining (ICDM). IEEE, 2017. pp. 913-918
@inproceedings{3a04a75801df4f459e86899d9999bc98,
title = "Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows",
abstract = "We address two important issues in causal discovery from nonstationary or heterogeneous data, where parameters associated with a causal structure may change over time or across data sets. First, we investigate how to efficiently estimate the {"}driving force{"} of the nonstationarity of a causal mechanism. That is, given a causal mechanism that varies over time or across data sets and whose qualitative structure is known, we aim to extract from data a low-dimensional and interpretable representation of the main components of the changes. For this purpose we develop a novel kernel embedding of nonstationary conditional distributions that does not rely on sliding windows. Second, the embedding also leads to a measure of dependence between the changes of causal modules that can be used to determine the directions of many causal arrows. We demonstrate the power of our methods with experiments on both synthetic and real data.",
author = "Biwei HUANG and Kun ZHANG and Jiji ZHANG and Ruben SANCHEZ-ROMERO and Clark GLYMOUR and Bernhard SCH{\"O}KOPF",
year = "2017",
language = "English",
publisher = "IEEE",
pages = "913--918",
booktitle = "2017 IEEE International Conference on Data Mining (ICDM)",

}

HUANG, B, ZHANG, K, ZHANG, J, SANCHEZ-ROMERO, R, GLYMOUR, C & SCHÖKOPF, B 2017, Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows. in 2017 IEEE International Conference on Data Mining (ICDM). IEEE, pp. 913-918, IEEE International Conference on Data Mining, New Orleans, United States, 18/11/17.

Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows. / HUANG, Biwei; ZHANG, Kun; ZHANG, Jiji; SANCHEZ-ROMERO, Ruben; GLYMOUR, Clark; SCHÖKOPF, Bernhard.

2017 IEEE International Conference on Data Mining (ICDM). IEEE, 2017. p. 913-918.

Research output: Book Chapters | Papers in Conference ProceedingsConference paper (refereed)Researchpeer-review

TY - GEN

T1 - Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows

AU - HUANG, Biwei

AU - ZHANG, Kun

AU - ZHANG, Jiji

AU - SANCHEZ-ROMERO, Ruben

AU - GLYMOUR, Clark

AU - SCHÖKOPF, Bernhard

PY - 2017

Y1 - 2017

N2 - We address two important issues in causal discovery from nonstationary or heterogeneous data, where parameters associated with a causal structure may change over time or across data sets. First, we investigate how to efficiently estimate the "driving force" of the nonstationarity of a causal mechanism. That is, given a causal mechanism that varies over time or across data sets and whose qualitative structure is known, we aim to extract from data a low-dimensional and interpretable representation of the main components of the changes. For this purpose we develop a novel kernel embedding of nonstationary conditional distributions that does not rely on sliding windows. Second, the embedding also leads to a measure of dependence between the changes of causal modules that can be used to determine the directions of many causal arrows. We demonstrate the power of our methods with experiments on both synthetic and real data.

AB - We address two important issues in causal discovery from nonstationary or heterogeneous data, where parameters associated with a causal structure may change over time or across data sets. First, we investigate how to efficiently estimate the "driving force" of the nonstationarity of a causal mechanism. That is, given a causal mechanism that varies over time or across data sets and whose qualitative structure is known, we aim to extract from data a low-dimensional and interpretable representation of the main components of the changes. For this purpose we develop a novel kernel embedding of nonstationary conditional distributions that does not rely on sliding windows. Second, the embedding also leads to a measure of dependence between the changes of causal modules that can be used to determine the directions of many causal arrows. We demonstrate the power of our methods with experiments on both synthetic and real data.

M3 - Conference paper (refereed)

SP - 913

EP - 918

BT - 2017 IEEE International Conference on Data Mining (ICDM)

PB - IEEE

ER -

HUANG B, ZHANG K, ZHANG J, SANCHEZ-ROMERO R, GLYMOUR C, SCHÖKOPF B. Behind Distribution Shift: Mining Driving Forces of Changes and Causal Arrows. In 2017 IEEE International Conference on Data Mining (ICDM). IEEE. 2017. p. 913-918