Recurrent neural networks for solving second-order cone programs

Chun Hsu Ko, Jein-Shan Chen, Ching Yu Yang

Research output: Contribution to journalArticle

22 Citations (Scopus)

Abstract

This paper proposes using the neural networks to efficiently solve the second-order cone programs (SOCP). To establish the neural networks, the SOCP is first reformulated as a second-order cone complementarity problem (SOCCP) with the Karush-Kuhn-Tucker conditions of the SOCP. The SOCCP functions, which transform the SOCCP into a set of nonlinear equations, are then utilized to design the neural networks. We propose two kinds of neural networks with the different SOCCP functions. The first neural network uses the Fischer-Burmeister function to achieve an unconstrained minimization with a merit function. We show that the merit function is a Lyapunov function and this neural network is asymptotically stable. The second neural network utilizes the natural residual function with the cone projection function to achieve low computation complexity. It is shown to be Lyapunov stable and converges globally to an optimal solution under some condition. The SOCP simulation results demonstrate the effectiveness of the proposed neural networks.

Original languageEnglish
Pages (from-to)3646-3653
Number of pages8
JournalNeurocomputing
Volume74
Issue number17
DOIs
Publication statusPublished - 2011 Oct 1

Fingerprint

Recurrent neural networks
Cones
Neural networks
Lyapunov functions
Set theory
Nonlinear equations

Keywords

  • Cone projection function
  • Fischer-Burmeister function
  • Lyapunov stable
  • Merit function
  • Neural network
  • SOCP

ASJC Scopus subject areas

  • Computer Science Applications
  • Cognitive Neuroscience
  • Artificial Intelligence

Cite this

Recurrent neural networks for solving second-order cone programs. / Ko, Chun Hsu; Chen, Jein-Shan; Yang, Ching Yu.

In: Neurocomputing, Vol. 74, No. 17, 01.10.2011, p. 3646-3653.

Research output: Contribution to journalArticle

Ko, Chun Hsu ; Chen, Jein-Shan ; Yang, Ching Yu. / Recurrent neural networks for solving second-order cone programs. In: Neurocomputing. 2011 ; Vol. 74, No. 17. pp. 3646-3653.
@article{6f2d1b660d14469db00bb1c28278b853,
title = "Recurrent neural networks for solving second-order cone programs",
abstract = "This paper proposes using the neural networks to efficiently solve the second-order cone programs (SOCP). To establish the neural networks, the SOCP is first reformulated as a second-order cone complementarity problem (SOCCP) with the Karush-Kuhn-Tucker conditions of the SOCP. The SOCCP functions, which transform the SOCCP into a set of nonlinear equations, are then utilized to design the neural networks. We propose two kinds of neural networks with the different SOCCP functions. The first neural network uses the Fischer-Burmeister function to achieve an unconstrained minimization with a merit function. We show that the merit function is a Lyapunov function and this neural network is asymptotically stable. The second neural network utilizes the natural residual function with the cone projection function to achieve low computation complexity. It is shown to be Lyapunov stable and converges globally to an optimal solution under some condition. The SOCP simulation results demonstrate the effectiveness of the proposed neural networks.",
keywords = "Cone projection function, Fischer-Burmeister function, Lyapunov stable, Merit function, Neural network, SOCP",
author = "Ko, {Chun Hsu} and Jein-Shan Chen and Yang, {Ching Yu}",
year = "2011",
month = "10",
day = "1",
doi = "10.1016/j.neucom.2011.07.009",
language = "English",
volume = "74",
pages = "3646--3653",
journal = "Neurocomputing",
issn = "0925-2312",
publisher = "Elsevier",
number = "17",

}

TY - JOUR

T1 - Recurrent neural networks for solving second-order cone programs

AU - Ko, Chun Hsu

AU - Chen, Jein-Shan

AU - Yang, Ching Yu

PY - 2011/10/1

Y1 - 2011/10/1

N2 - This paper proposes using the neural networks to efficiently solve the second-order cone programs (SOCP). To establish the neural networks, the SOCP is first reformulated as a second-order cone complementarity problem (SOCCP) with the Karush-Kuhn-Tucker conditions of the SOCP. The SOCCP functions, which transform the SOCCP into a set of nonlinear equations, are then utilized to design the neural networks. We propose two kinds of neural networks with the different SOCCP functions. The first neural network uses the Fischer-Burmeister function to achieve an unconstrained minimization with a merit function. We show that the merit function is a Lyapunov function and this neural network is asymptotically stable. The second neural network utilizes the natural residual function with the cone projection function to achieve low computation complexity. It is shown to be Lyapunov stable and converges globally to an optimal solution under some condition. The SOCP simulation results demonstrate the effectiveness of the proposed neural networks.

AB - This paper proposes using the neural networks to efficiently solve the second-order cone programs (SOCP). To establish the neural networks, the SOCP is first reformulated as a second-order cone complementarity problem (SOCCP) with the Karush-Kuhn-Tucker conditions of the SOCP. The SOCCP functions, which transform the SOCCP into a set of nonlinear equations, are then utilized to design the neural networks. We propose two kinds of neural networks with the different SOCCP functions. The first neural network uses the Fischer-Burmeister function to achieve an unconstrained minimization with a merit function. We show that the merit function is a Lyapunov function and this neural network is asymptotically stable. The second neural network utilizes the natural residual function with the cone projection function to achieve low computation complexity. It is shown to be Lyapunov stable and converges globally to an optimal solution under some condition. The SOCP simulation results demonstrate the effectiveness of the proposed neural networks.

KW - Cone projection function

KW - Fischer-Burmeister function

KW - Lyapunov stable

KW - Merit function

KW - Neural network

KW - SOCP

UR - http://www.scopus.com/inward/record.url?scp=80052937770&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=80052937770&partnerID=8YFLogxK

U2 - 10.1016/j.neucom.2011.07.009

DO - 10.1016/j.neucom.2011.07.009

M3 - Article

VL - 74

SP - 3646

EP - 3653

JO - Neurocomputing

JF - Neurocomputing

SN - 0925-2312

IS - 17

ER -