Predicting the Solution Time of Branch-and-Bound Algorithms for Mixed-Integer Programs

@article{zaltn2011PredictingTS,
  title={Predicting the Solution Time of Branch-and-Bound Algorithms for Mixed-Integer Programs},
  author={Osman Y. {\"O}zaltın and Brady Hunsaker and Andrew J. Schaefer},
  journal={INFORMS J. Comput.},
  year={2011},
  volume={23},
  pages={392-403}
}
The most widely used progress measure for branch-and-bound (B&B) algorithms when solving mixed-integer programs (MIPs) is the MIP gap. We introduce a new progress measure that is often much smoother than the MIP gap. We propose a double exponential smoothing technique to predict the solution time of B&B algorithms and evaluate the prediction method using three MIP solvers. Our computational experiments show that accurate predictions of the solution time are possible, even in the early stages of… 
Estimating the Size of Branch-And-Bound Trees
This paper investigates the estimation of the size of Branch-and-Bound (B&B) trees for solving mixed-integer programs. We first prove that the size of the B&B tree cannot be approximated within a
Clairvoyant Restarts in Branch-and-Bound Search Using Online Tree-Size Estimation
TLDR
This work designs a restart strategy for Mixed-Integer Programming (MIP) solvers that decides whether to restart the search based on the current estimate of the number of remaining nodes in the tree, and refers to this type of algorithm as clairvoyant.
On estimating workload in interval branch-and-bound global optimization algorithms
TLDR
This work study new methods to estimate the left-over value based on the observed amount of pruning that provide information about the remaining running time of the algorithm and the required computational resources.
Parallel Solvers for Mixed Integer Linear Programming
TLDR
This article introduces parallel mixed integer linear programming (MILP) solvers and summarizes current existing parallel MILP solver architectures.
Estimating the size of search trees by sampling with domain knowledge
TLDR
It is shown how recently-defined abstract models of the Branch & Bound algorithm can be used to obtain information on how the nodes are distributed in B&B search trees through a sampling algorithm given by Knuth that estimates the size of a search tree.
The SCIP Optimization Suite 5.0
TLDR
New features and enhanced algorithms made available in version 5.0 of the SCIP Optimization Suite, in particular for the LP solver SoPlex, the Steiner tree solver SCIP-Jack, the MISDP solverSCIP-SDP, and the parallelization framework UG are described.
Parallel Solvers for Mixed Integer Linear Optimization
TLDR
This chapter provides an overview of the current state of the art with respect to solution of mixed integer linear optimization problems (MILPs) in parallel and describes how existing parallel MILP solvers can be classified according to various properties of the underlying algorithm.
Self-split parallelization for Mixed-Integer Linear Programming
SelfSplit is a simple static mechanism to convert a sequential tree-search code into a parallel one. In this paradigm, tree-search is distributed among a set of identical workers, each of which is
Could we use a million cores to solve an integer program?
TLDR
The opportunities and challenges that will arise when trying to utilize this vast computing power to solve a single integer linear optimization problem are investigated and whether best practices in sequential solution of ILPs will be effective in massively parallel environments is raised.
SelfSplit parallelization for mixed-integer linear programming
TLDR
Computational results show that SelfSplit, in spite of its simplicity, can achieve good speedups even in the MILP context.
...
1
2
...

References

SHOWING 1-10 OF 21 REFERENCES
Early Estimates of the Size of Branch-and-Bound Trees
TLDR
The time needed to solve mixed-integer-programming problems by branch and bound can be roughly predicted early in the solution process, based on analyzing the partial tree resulting from running the algorithm for a short period of time and predicting the shape of the whole tree.
Heuristic Sampling: A Method for Predicting the Performance of Tree Searching Programs
  • Pang C. Chen
  • Mathematics, Computer Science
    SIAM J. Comput.
  • 1992
TLDR
A new method based on a stratified sampling approach, which is called heuristic sampling, is a generalization of Knuth’s original algorithm for estimating the efficiency of backtrack programs and can produce significantly more accurate cost estimates for commonly used tree search algorithms.
Estimating the efficiency of backtrack programs.
One of the chief difficulties associated with the so-called backtracking technique for combinatorial problems has been our inability to predict the efficiency of a given algorithm, or to compare the
Miplib 2003
TLDR
This paper reports on the fourth version of the Mixed Integer Programming Library, and presents an overview of the 27 new problems and statistical data for all 60 instances that became too easy.
Estimating Search Tree Size
TLDR
This work proposes two new online methods for estimating the size of a backtracking search tree based on a weighted sample of the branches visited by chronological backtracking and compares them against an old method due to Knuth based on random probing.
Tree Size by Partial Backtracking
  • P. Purdom
  • Mathematics, Computer Science
    SIAM J. Comput.
  • 1978
TLDR
The efficiency of Knuth's method can be greatly improved by occasionally following more than one path from a node, which results in an improvement which increases exponentially with the height of the tree.
Forecasting for Inventory Control with Exponential Smoothing
Exponential smoothing, often used for sales forecasting in inventory control, has always been rationalized in terms of statistical models that possess errors with constant variances. It is shown in
FORECASTING WITH EXPONENTIAL SMOOTHING: SOME GUIDELINES FOR MODEL SELECTION
Despite the general acceptance of exponential smoothing, the choice of a specific smoothing model is often a difficult problem. Previous research involving smoothing-model comparisons and the
Double Exponential Smoothing: An Alternative to Kalman Filter-Based Predictive Tracking
TLDR
Novel algorithms for predictive tracking of user position and orientation based on double exponential smoothing are presented, which are faster, easier to implement, and perform equivalently to the Kalman and extended Kalman filtering predictors.
An introduction to Kalman filter
3 A deterministic construction of the Kalman filter 7 3.1 The quadratic criterion to be minimized in the absence of process control noise and the corresponding solution . . . . . . . . . . . . . . .
...
1
2
3
...