Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming
description
Transcript of Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming
![Page 1: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/1.jpg)
Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer
Programming
Sanjay MehrotraDepartment of IE/MS
Northwestern University
Evanston, IL 60208
Joint Work with
Zhifeng Liand
Huan-yuan Sheng
![Page 2: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/2.jpg)
Organization
Introduction and Review Adjoint Lattice Theoretical Results Computational Results Conclusions
![Page 3: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/3.jpg)
Problem Formulation
![Page 4: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/4.jpg)
Approaches to Branching
Branching StrategiesBranching on variablesMost infeasible branchingStrong branchingNested Cluster BranchingPseudo-cost branchingBranching on constraints
Branching on variables may not be efficient for solving general MIP.
![Page 5: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/5.jpg)
An Example
![Page 6: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/6.jpg)
Example (continued)
![Page 7: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/7.jpg)
An Example (continued)
![Page 8: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/8.jpg)
Branching on Hyperplane Algorithm(s)
Lenstra’s Algorithm
![Page 9: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/9.jpg)
Algorithm Continued…
Generalized Basis
Reduction in Lovasz-
Scarf Algorithm
![Page 10: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/10.jpg)
Branching on hyperplane Algorithm
Step 2:
Step 3:
![Page 11: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/11.jpg)
Pervious Work
Lenstra [83] proposed algorithm branching on hyperplane Grötschel, Lovász and Schrijver [84] used ellipsoidal approximation directly. Lovász and Scarf [92] developed Generalized Basis Reduction (GBR) algorithm
that does not require ellipsoidal approximation and works with the original model, however, assumes full dimensionality.
Cook, et. al. [93] implemented GBR algorithm for some hard network design problems.
Wang [97] implemented GBR algorithm for LP and NLP (<100 integer variables).
Aardal, Hurkens, and Lenstra [98,00], Aardal et. al. [00], Aardal and Lenstra [02], proposed a reformulation technique and solved some hard equality constrained integer knapsack and market split problems using this reformulation.
Owen and Mehrotra [02] computationally showed that good branching disjunctions are available using heuristics for smaller problems in the MIPLIB testset.
Gao and Zhang [02] implemented Lenstra's algorithm and tested an interior point algorithm for finding the maximum volume ellipsoid.
![Page 12: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/12.jpg)
Issues
Dimension reduction… hmm…• Don’t know what to do when continuous variables are present
Continuous variables are “projected” out
• How to do this for the more general convex problem?• It is just nice to work in the original space!
Basis Reduction at every node…• Finding a feasible solution or • finding good branching directions
Ellipsoidal Approximation• We should use the modern technologies!
Feasibility problem vs. Optimization Problem• Disjunctive branching (instead of branching on hyperplanes)
![Page 13: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/13.jpg)
Matter of Definitions
![Page 14: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/14.jpg)
Adjoint Lattice
![Page 15: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/15.jpg)
Branching Hyperplane finding Problem
![Page 16: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/16.jpg)
Find An integral Adjoint Basis
![Page 17: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/17.jpg)
How to Solve Minimum Width Problems
![Page 18: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/18.jpg)
Lovasz and Scarf Basis
![Page 19: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/19.jpg)
Mixed Integer Programming
Feasibility Mixed Integer Linear Program (FMILP) is to
![Page 20: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/20.jpg)
Mixed Integer Problem
![Page 21: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/21.jpg)
Generalized B&B Method
A growing search tree
IMPACT (Integrated Mathematical Programming Advanced Computational Tool) implementation of GBB.
![Page 22: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/22.jpg)
Knapsack Test Problems
![Page 23: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/23.jpg)
Numerical Results on Hard Knapsack Problems
![Page 24: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/24.jpg)
Market Split Problems
![Page 25: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/25.jpg)
Market Split Problems
![Page 26: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/26.jpg)
Larger Market Split Problems
![Page 27: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/27.jpg)
# of LLL iterations and GBB Tree Size
![Page 28: Computational Experiences with Branching on Hyperplane Algorithm for Mixed Integer Programming](https://reader038.fdocuments.us/reader038/viewer/2022110103/56814c3a550346895db940c6/html5/thumbnails/28.jpg)
Conclusions Work in progress
It is possible to develop stable codes for Dense Difficult Market Split problems using Adjoint Lattice Basis.
Increasing the swap threshold in LLL algorithm reduces the total number of branches, but this is not consistent beyond .99
GBR algorithm in Lova’sz-Scarf branching on hyperplane algorithm is sensitive to optimality precision.