A Deterministic Annealing Approach for Parsimonious Design of Piecewise Regression Models

Ajit V. Rao; David J. Miller; Kenneth Rose; Alien Gersho

doi:10.1109/34.748824

A Deterministic Annealing Approach for Parsimonious Design of Piecewise Regression Models

Ajit V. Rao, David J. Miller, Kenneth Rose, Alien Gersho

Electrical Engineering

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

47 Scopus citations

Abstract

A new learning algorithm is proposed for piecewise regression modeling. It employs the technique of deterministic annealing to design space partition regression functions. While the performance of traditional space partition regression functions such as CART and MARS is limited by a simple tree-structured partition and by a hierarchical approach for design the deterministic annealing algorithm enables the joint optimization of a more powerful piecewise structure based on a Voronoi partition. The new method is demonstrated to achieve consistent performance improvements over regular CART as well as over its extension to allow arbitrary hyperplane boundaries. Comparison tests on several benchmark data sets from the regression literature are provided.

Original language	English (US)
Pages (from-to)	159-173
Number of pages	15
Journal	IEEE Transactions on Pattern Analysis and Machine Intelligence
Volume	21
Issue number	2
DOIs	https://doi.org/10.1109/34.748824
State	Published - 1999

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition
Computational Theory and Mathematics
Artificial Intelligence
Applied Mathematics

Access to Document

10.1109/34.748824

Cite this

@article{10a29f4d66ac4ef7a7b4e4e29ac291a8,

title = "A Deterministic Annealing Approach for Parsimonious Design of Piecewise Regression Models",

abstract = "A new learning algorithm is proposed for piecewise regression modeling. It employs the technique of deterministic annealing to design space partition regression functions. While the performance of traditional space partition regression functions such as CART and MARS is limited by a simple tree-structured partition and by a hierarchical approach for design the deterministic annealing algorithm enables the joint optimization of a more powerful piecewise structure based on a Voronoi partition. The new method is demonstrated to achieve consistent performance improvements over regular CART as well as over its extension to allow arbitrary hyperplane boundaries. Comparison tests on several benchmark data sets from the regression literature are provided.",

author = "Rao, {Ajit V.} and Miller, {David J.} and Kenneth Rose and Alien Gersho",

note = "Funding Information: This work was supported in part by the National Science Foundation under grant no. NCR-9314335, the University of California MICRO program, ACT Networks, Advanced Computer Communications, Cisco Systems, DSP Group, DSP Software Engineering, Fujitsu Laboratories of America, General Electric, Hughes Electronics, Intel, Nokia Mobile Phones, Qualcomm, Rockwell International, and Texas Instruments. David Miller was supported by NSF Career Award NSF IRI-9624870. Most of this work was done when A.V. Rao and D.J. Miller were graduate students in the De- partment of Electrical and Computer Engineering, University of California, Santa Barbara.",

year = "1999",

doi = "10.1109/34.748824",

language = "English (US)",

volume = "21",

pages = "159--173",

journal = "IEEE Transactions on Pattern Analysis and Machine Intelligence",

issn = "0162-8828",

publisher = "IEEE Computer Society",

number = "2",

}

TY - JOUR

T1 - A Deterministic Annealing Approach for Parsimonious Design of Piecewise Regression Models

AU - Rao, Ajit V.

AU - Miller, David J.

AU - Rose, Kenneth

AU - Gersho, Alien

N1 - Funding Information: This work was supported in part by the National Science Foundation under grant no. NCR-9314335, the University of California MICRO program, ACT Networks, Advanced Computer Communications, Cisco Systems, DSP Group, DSP Software Engineering, Fujitsu Laboratories of America, General Electric, Hughes Electronics, Intel, Nokia Mobile Phones, Qualcomm, Rockwell International, and Texas Instruments. David Miller was supported by NSF Career Award NSF IRI-9624870. Most of this work was done when A.V. Rao and D.J. Miller were graduate students in the De- partment of Electrical and Computer Engineering, University of California, Santa Barbara.

PY - 1999

Y1 - 1999

N2 - A new learning algorithm is proposed for piecewise regression modeling. It employs the technique of deterministic annealing to design space partition regression functions. While the performance of traditional space partition regression functions such as CART and MARS is limited by a simple tree-structured partition and by a hierarchical approach for design the deterministic annealing algorithm enables the joint optimization of a more powerful piecewise structure based on a Voronoi partition. The new method is demonstrated to achieve consistent performance improvements over regular CART as well as over its extension to allow arbitrary hyperplane boundaries. Comparison tests on several benchmark data sets from the regression literature are provided.

AB - A new learning algorithm is proposed for piecewise regression modeling. It employs the technique of deterministic annealing to design space partition regression functions. While the performance of traditional space partition regression functions such as CART and MARS is limited by a simple tree-structured partition and by a hierarchical approach for design the deterministic annealing algorithm enables the joint optimization of a more powerful piecewise structure based on a Voronoi partition. The new method is demonstrated to achieve consistent performance improvements over regular CART as well as over its extension to allow arbitrary hyperplane boundaries. Comparison tests on several benchmark data sets from the regression literature are provided.

UR - http://www.scopus.com/inward/record.url?scp=0033075499&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=0033075499&partnerID=8YFLogxK

U2 - 10.1109/34.748824

DO - 10.1109/34.748824

M3 - Article

AN - SCOPUS:0033075499

SN - 0162-8828

VL - 21

SP - 159

EP - 173

JO - IEEE Transactions on Pattern Analysis and Machine Intelligence

JF - IEEE Transactions on Pattern Analysis and Machine Intelligence

IS - 2

ER -

A Deterministic Annealing Approach for Parsimonious Design of Piecewise Regression Models

Abstract

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this