MINLP formulations for continuous piecewise linear function fitting

Noam Goldberg; Steffen Rebennack; Youngdae Kim; Vitaliy Krasko; Sven Leyffer

doi:10.1007/s10589-021-00268-5

MINLP formulations for continuous piecewise linear function fitting

Noam Goldberg, Steffen Rebennack, Youngdae Kim, Vitaliy Krasko, Sven Leyffer

Department of Management

Research output: Contribution to journal › Article › peer-review

5 Scopus citations

Abstract

We consider a nonconvex mixed-integer nonlinear programming (MINLP) model proposed by Goldberg et al. (Comput Optim Appl 58:523–541, 2014. https://doi.org/10.1007/s10589-014-9647-y) for piecewise linear function fitting. We show that this MINLP model is incomplete and can result in a piecewise linear curve that is not the graph of a function, because it misses a set of necessary constraints. We provide two counterexamples to illustrate this effect, and propose three alternative models that correct this behavior. We investigate the theoretical relationship between these models and evaluate their computational performance.

Original language	English
Pages (from-to)	223-233
Number of pages	11
Journal	Computational Optimization and Applications
Volume	79
Issue number	1
DOIs	https://doi.org/10.1007/s10589-021-00268-5
State	Published - May 2021

Bibliographical note

Publisher Copyright:
© 2021, The Author(s).

Funding

Funders	Funder number
Deutsche Forschungsgemeinschaft	445857709

Keywords

Branch-and-bound
Linear spline regression
Mixed-integer nonlinear program
Reformulation

Access to Document

10.1007/s10589-021-00268-5

Cite this

@article{304a33ad36624d92b724820802e1f9b9,

title = "MINLP formulations for continuous piecewise linear function fitting",

abstract = "We consider a nonconvex mixed-integer nonlinear programming (MINLP) model proposed by Goldberg et al. (Comput Optim Appl 58:523–541, 2014. https://doi.org/10.1007/s10589-014-9647-y) for piecewise linear function fitting. We show that this MINLP model is incomplete and can result in a piecewise linear curve that is not the graph of a function, because it misses a set of necessary constraints. We provide two counterexamples to illustrate this effect, and propose three alternative models that correct this behavior. We investigate the theoretical relationship between these models and evaluate their computational performance.",

keywords = "Branch-and-bound, Linear spline regression, Mixed-integer nonlinear program, Reformulation",

author = "Noam Goldberg and Steffen Rebennack and Youngdae Kim and Vitaliy Krasko and Sven Leyffer",

note = "Publisher Copyright: {\textcopyright} 2021, The Author(s).",

year = "2021",

month = may,

doi = "10.1007/s10589-021-00268-5",

language = "אנגלית",

volume = "79",

pages = "223--233",

journal = "Computational Optimization and Applications",

issn = "0926-6003",

publisher = "Springer Netherlands",

number = "1",

}

TY - JOUR

T1 - MINLP formulations for continuous piecewise linear function fitting

AU - Goldberg, Noam

AU - Rebennack, Steffen

AU - Kim, Youngdae

AU - Krasko, Vitaliy

AU - Leyffer, Sven

PY - 2021/5

Y1 - 2021/5

N2 - We consider a nonconvex mixed-integer nonlinear programming (MINLP) model proposed by Goldberg et al. (Comput Optim Appl 58:523–541, 2014. https://doi.org/10.1007/s10589-014-9647-y) for piecewise linear function fitting. We show that this MINLP model is incomplete and can result in a piecewise linear curve that is not the graph of a function, because it misses a set of necessary constraints. We provide two counterexamples to illustrate this effect, and propose three alternative models that correct this behavior. We investigate the theoretical relationship between these models and evaluate their computational performance.

AB - We consider a nonconvex mixed-integer nonlinear programming (MINLP) model proposed by Goldberg et al. (Comput Optim Appl 58:523–541, 2014. https://doi.org/10.1007/s10589-014-9647-y) for piecewise linear function fitting. We show that this MINLP model is incomplete and can result in a piecewise linear curve that is not the graph of a function, because it misses a set of necessary constraints. We provide two counterexamples to illustrate this effect, and propose three alternative models that correct this behavior. We investigate the theoretical relationship between these models and evaluate their computational performance.

KW - Branch-and-bound

KW - Linear spline regression

KW - Mixed-integer nonlinear program

KW - Reformulation

UR - http://www.scopus.com/inward/record.url?scp=85102374580&partnerID=8YFLogxK

U2 - 10.1007/s10589-021-00268-5

DO - 10.1007/s10589-021-00268-5

M3 - ???researchoutput.researchoutputtypes.contributiontojournal.article???

AN - SCOPUS:85102374580

SN - 0926-6003

VL - 79

SP - 223

EP - 233

JO - Computational Optimization and Applications

JF - Computational Optimization and Applications

IS - 1

ER -

MINLP formulations for continuous piecewise linear function fitting

Abstract

Bibliographical note

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this