Change surfaces for expressive multidimensional changepoints and counterfactual prediction

William Herlands, Daniel Neill, Hannes Nickisch, Andrew Gordon Wilson

Research output: Contribution to journalArticle

Abstract

Identifying changes in model parameters is fundamental in machine learning and statistics. However, standard changepoint models are limited in expressiveness, often addressing unidimensional problems and assuming instantaneous changes. We introduce change surfaces as a multidimensional and highly expressive generalization of changepoints. We provide a model-agnostic formalization of change surfaces, illustrating how they can provide variable, heterogeneous, and non-monotonic rates of change across multiple dimensions. Additionally, we show how change surfaces can be used for counterfactual prediction. As a concrete instantiation of the change surface framework, we develop Gaussian Process Change Surfaces (GPCS). We demonstrate counterfactual prediction with Bayesian posterior mean and credible sets, as well as massive scalability by introducing novel methods for additive nonseparable kernels. Using two large spatio-temporal datasets we employ GPCS to discover and characterize complex changes that can provide scientific and policy relevant insights. Specifically, we analyze twentieth century measles incidence across the United States and discover previously unknown heterogeneous changes after the introduction of the measles vaccine. Additionally, we apply the model to requests for lead testing kits in New York City, discovering distinct spatial and demographic patterns.

Original languageEnglish (US)
JournalJournal of Machine Learning Research
Volume20
StatePublished - Jun 1 2019

Fingerprint

Change Point
Prediction
Gaussian Process
Change-point Model
Posterior Mean
Vaccines
Rate of change
Vaccine
Nonseparable
Expressiveness
Formalization
Instantaneous
Learning systems
Scalability
Standard Model
Incidence
Machine Learning
Lead
Statistics
Model

Keywords

  • Change surface
  • Changepoint
  • Counterfactual
  • Gaussian process
  • Kernel method
  • Scalable inference

ASJC Scopus subject areas

  • Software
  • Control and Systems Engineering
  • Statistics and Probability
  • Artificial Intelligence

Cite this

Change surfaces for expressive multidimensional changepoints and counterfactual prediction. / Herlands, William; Neill, Daniel; Nickisch, Hannes; Wilson, Andrew Gordon.

In: Journal of Machine Learning Research, Vol. 20, 01.06.2019.

Research output: Contribution to journalArticle

@article{42451da2f00e454ba35d2074527cd97d,
title = "Change surfaces for expressive multidimensional changepoints and counterfactual prediction",
abstract = "Identifying changes in model parameters is fundamental in machine learning and statistics. However, standard changepoint models are limited in expressiveness, often addressing unidimensional problems and assuming instantaneous changes. We introduce change surfaces as a multidimensional and highly expressive generalization of changepoints. We provide a model-agnostic formalization of change surfaces, illustrating how they can provide variable, heterogeneous, and non-monotonic rates of change across multiple dimensions. Additionally, we show how change surfaces can be used for counterfactual prediction. As a concrete instantiation of the change surface framework, we develop Gaussian Process Change Surfaces (GPCS). We demonstrate counterfactual prediction with Bayesian posterior mean and credible sets, as well as massive scalability by introducing novel methods for additive nonseparable kernels. Using two large spatio-temporal datasets we employ GPCS to discover and characterize complex changes that can provide scientific and policy relevant insights. Specifically, we analyze twentieth century measles incidence across the United States and discover previously unknown heterogeneous changes after the introduction of the measles vaccine. Additionally, we apply the model to requests for lead testing kits in New York City, discovering distinct spatial and demographic patterns.",
keywords = "Change surface, Changepoint, Counterfactual, Gaussian process, Kernel method, Scalable inference",
author = "William Herlands and Daniel Neill and Hannes Nickisch and Wilson, {Andrew Gordon}",
year = "2019",
month = "6",
day = "1",
language = "English (US)",
volume = "20",
journal = "Journal of Machine Learning Research",
issn = "1532-4435",
publisher = "Microtome Publishing",

}

TY - JOUR

T1 - Change surfaces for expressive multidimensional changepoints and counterfactual prediction

AU - Herlands, William

AU - Neill, Daniel

AU - Nickisch, Hannes

AU - Wilson, Andrew Gordon

PY - 2019/6/1

Y1 - 2019/6/1

N2 - Identifying changes in model parameters is fundamental in machine learning and statistics. However, standard changepoint models are limited in expressiveness, often addressing unidimensional problems and assuming instantaneous changes. We introduce change surfaces as a multidimensional and highly expressive generalization of changepoints. We provide a model-agnostic formalization of change surfaces, illustrating how they can provide variable, heterogeneous, and non-monotonic rates of change across multiple dimensions. Additionally, we show how change surfaces can be used for counterfactual prediction. As a concrete instantiation of the change surface framework, we develop Gaussian Process Change Surfaces (GPCS). We demonstrate counterfactual prediction with Bayesian posterior mean and credible sets, as well as massive scalability by introducing novel methods for additive nonseparable kernels. Using two large spatio-temporal datasets we employ GPCS to discover and characterize complex changes that can provide scientific and policy relevant insights. Specifically, we analyze twentieth century measles incidence across the United States and discover previously unknown heterogeneous changes after the introduction of the measles vaccine. Additionally, we apply the model to requests for lead testing kits in New York City, discovering distinct spatial and demographic patterns.

AB - Identifying changes in model parameters is fundamental in machine learning and statistics. However, standard changepoint models are limited in expressiveness, often addressing unidimensional problems and assuming instantaneous changes. We introduce change surfaces as a multidimensional and highly expressive generalization of changepoints. We provide a model-agnostic formalization of change surfaces, illustrating how they can provide variable, heterogeneous, and non-monotonic rates of change across multiple dimensions. Additionally, we show how change surfaces can be used for counterfactual prediction. As a concrete instantiation of the change surface framework, we develop Gaussian Process Change Surfaces (GPCS). We demonstrate counterfactual prediction with Bayesian posterior mean and credible sets, as well as massive scalability by introducing novel methods for additive nonseparable kernels. Using two large spatio-temporal datasets we employ GPCS to discover and characterize complex changes that can provide scientific and policy relevant insights. Specifically, we analyze twentieth century measles incidence across the United States and discover previously unknown heterogeneous changes after the introduction of the measles vaccine. Additionally, we apply the model to requests for lead testing kits in New York City, discovering distinct spatial and demographic patterns.

KW - Change surface

KW - Changepoint

KW - Counterfactual

KW - Gaussian process

KW - Kernel method

KW - Scalable inference

UR - http://www.scopus.com/inward/record.url?scp=85072632086&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85072632086&partnerID=8YFLogxK

M3 - Article

AN - SCOPUS:85072632086

VL - 20

JO - Journal of Machine Learning Research

JF - Journal of Machine Learning Research

SN - 1532-4435

ER -