A mixed-integer quadratic programming solver based on GPU

Solving the mixed-integer quadratic programming (MIQP) problem is often required in many practical applications. But the existing solvers always encounter the contradiction between high precision and low time consumption. To solve this problem, this paper designs a new MIQP solver by developing a pa...

Full description

Saved in:
Bibliographic Details
Published in:2015 34th Chinese Control Conference (CCC) pp. 2686 - 2691
Main Authors: Xi, Wang, Dewei, Li, Yugeng, Xi
Format: Conference Proceeding Journal Article
Language:English
Published: Technical Committee on Control Theory, Chinese Association of Automation 01.07.2015
Subjects:
Algorithm design and analysis
Algorithms
Approximation algorithms
Branch and Bound Algorithm
Computational efficiency
Conferences
GPU
Graphics processing units
Instruction sets
Mix-integer Quadratic Programming
Neural networks
Neurons
Parallel Structure
Quadratic programming
Searching
Solvers
Statistical tests
Strategy
ISSN:1934-1768
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Solving the mixed-integer quadratic programming (MIQP) problem is often required in many practical applications. But the existing solvers always encounter the contradiction between high precision and low time consumption. To solve this problem, this paper designs a new MIQP solver by developing a parallel branch-and-bound algorithm utilizing multi-point radiation based on the multithreading parallel structure of GPU. This solver can obtain the global optimal solution by inheriting the advantages of the classical branch-and-bound algorithm. To increase the efficiency of the MIQP solver, for the quadratic programming (QP) to be solved each time during iteration, we fully use the massive parallelism of GPU and adopt the discrete-time simplified dual neural network. The idea of multi-point radiation is used to simultaneously generate multiple search branches to improve the search efficiency. These strategies enhance the throughput and the degree of parallelization. The high computational efficiency of the proposed MIQP solver is verified by test results with solving time statistics for multiple examples.
Bibliography:ObjectType-Article-2
SourceType-Scholarly Journals-1
ObjectType-Conference-1
ObjectType-Feature-3
content type line 23
SourceType-Conference Papers & Proceedings-2
ISSN:1934-1768
DOI:10.1109/ChiCC.2015.7260048