Model Evaluation for Geospatial Problems

Jing Wang; Tyler Hallman; Laurel Hopkins; John Burns Kilbride; W Douglas Robinson; Rebecca Hutchinson

Model Evaluation for Geospatial Problems

Jing Wang
, Tyler Hallman
, Laurel Hopkins
, John Burns Kilbride
, W Douglas Robinson
, Rebecca Hutchinson

Ysgol Gwyddorau Amgylcheddol a Naturiol

Marine Mammal Institute, Hatfield Marine Science Center, Oregon State University, Newport, Oregon

Allbwn ymchwil: Cyfraniad at gynhadledd › Papur › adolygiad gan gymheiriaid

Crynodeb

Geospatial problems often involve spatial autocorrelation and covariate shift, which
violate the independent, identically distributed assumption underlying standard
cross-validation. In this work, we establish a theoretical criterion for unbiased cross-validation, introduce a preliminary categorization framework to guide practitioners in choosing suitable cross-validation strategies for geospatial problems, reconcile conflicting recommendations on best practices, and develop a novel, straightforward method with both theoretical guarantees and empirical success.

Iaith wreiddiol	Saesneg
Statws	Cyhoeddwyd - 30 Rhag 2023
Digwyddiad	2023 NeurIPS Workshop on Computational Sustainability: Pitfalls and Promises from Theory to Deployment - New Orleans Hyd: 15 Rhag 2023 → …

Cynhadledd

Cynhadledd	2023 NeurIPS Workshop on Computational Sustainability: Pitfalls and Promises from Theory to Deployment
Dinas	New Orleans
Cyfnod	15/12/23 → …

Dyfynnu hyn

@conference{ce0e90dc30994a3badb43e47d5b7d42b,

title = "Model Evaluation for Geospatial Problems",

abstract = "Geospatial problems often involve spatial autocorrelation and covariate shift, which violate the independent, identically distributed assumption underlying standard cross-validation. In this work, we establish a theoretical criterion for unbiased cross-validation, introduce a preliminary categorization framework to guide practitioners in choosing suitable cross-validation strategies for geospatial problems, reconcile conflicting recommendations on best practices, and develop a novel, straightforward method with both theoretical guarantees and empirical success.",

author = "Jing Wang and Tyler Hallman and Laurel Hopkins and Kilbride, \{John Burns\} and Robinson, \{W Douglas\} and Rebecca Hutchinson",

year = "2023",

month = dec,

day = "30",

language = "English",

note = "2023 NeurIPS Workshop on Computational Sustainability: Pitfalls and Promises from Theory to Deployment ; Conference date: 15-12-2023",

}

TY - CONF

T1 - Model Evaluation for Geospatial Problems

AU - Wang, Jing

AU - Hallman, Tyler

AU - Hopkins, Laurel

AU - Kilbride, John Burns

AU - Robinson, W Douglas

AU - Hutchinson, Rebecca

PY - 2023/12/30

Y1 - 2023/12/30

N2 - Geospatial problems often involve spatial autocorrelation and covariate shift, which violate the independent, identically distributed assumption underlying standard cross-validation. In this work, we establish a theoretical criterion for unbiased cross-validation, introduce a preliminary categorization framework to guide practitioners in choosing suitable cross-validation strategies for geospatial problems, reconcile conflicting recommendations on best practices, and develop a novel, straightforward method with both theoretical guarantees and empirical success.

AB - Geospatial problems often involve spatial autocorrelation and covariate shift, which violate the independent, identically distributed assumption underlying standard cross-validation. In this work, we establish a theoretical criterion for unbiased cross-validation, introduce a preliminary categorization framework to guide practitioners in choosing suitable cross-validation strategies for geospatial problems, reconcile conflicting recommendations on best practices, and develop a novel, straightforward method with both theoretical guarantees and empirical success.

M3 - Paper

T2 - 2023 NeurIPS Workshop on Computational Sustainability: Pitfalls and Promises from Theory to Deployment

Y2 - 15 December 2023

ER -

Model Evaluation for Geospatial Problems

Crynodeb

Cynhadledd

Ôl bys

Dyfynnu hyn