Algorithm portfolios

@article{Gomes2001AlgorithmP,
  title={Algorithm portfolios},
  author={Carla P. Gomes and Bart Selman},
  journal={Artif. Intell.},
  year={2001},
  volume={126},
  pages={43-62}
}

Evolution of Algorithm Portfolio Methods for Combinatorial Search and Optimization Strategies

TLDR
The development of the principal approaches that leverage machine learning to accelerate the solution search of modern solvers are illustrated and the basic notions necessary to understand the process that enabled these techniques to achieve their amazing results are presented.

On Maximum Speedup Ratio of Restart Algorithm Portfolios

TLDR
This paper proves that in the best case the mixed algorithm portfolio may perform approximately up to 1.58 times faster than the best single algorithm portfolio, and estimates the computational value of mixing randomized restart algorithms with different properties.

Learning parallel portfolios of algorithms

TLDR
This work presents an effective method for finding a PPA in which the share of processor time allocated to each algorithm is fixed and presents bounds on the performance of the PPA over random instances and evaluates the performance empirically on a collection of 23 state-of-the-art SAT algorithms.

Portfolio approaches in constraint programming

Recent research has shown that the performance of a single, arbitrarily efficient algorithm can be significantly outperformed by using a portfolio of —possibly on-average slower— algorithms. Within

Search Portfolio with Sharing

TLDR
A new search framework, called Search Portfolio with Sharing (SP-S), which uses multiple algorithms to explore a given state-space in an integrated manner, seamlessly combining the partial solutions, while preserving the constraints/characteristics of the candidate algorithms.

Statistical Models of Multistart Randomized Heuristic Search Performance

TLDR
This paper presents some of the analysis of potential models, including goodness of fit tests for several possible assumptions the authors can make about the distribution of the quality of solutions obtained by VBSS over its allocated set of restarts, and leads to the selection of the Generalized Extreme Value Distribution for models of randomized heuristic performance.

Date of acceptance Grade Instructor Algorithm portfolios in constraint solving

TLDR
SATzilla is presented, a generic process for constructing an algorithm portfolio that utilizes so called empirical hardness models to predict an algorithm's running time on given instance.

Real-time solving of computationally hard problems using optimal algorithm portfolios

TLDR
The first systematic empirical evaluation of the relative success of various known stochastic-search algorithms in coping with a hard combinatorial optimization problems under a very short and fixed timeout is given.

Designing and Comparing Multiple Portfolios of Parameter Configurations for Online Algorithm Selection

TLDR
This paper employs a Design of Experiments (DOE) approach to determine a promising range of values for each parameter of an algorithm, which would be used within two online Algorithm Selection approaches for solving different instances of a given combinatorial optimization problem effectively.

Parallel Strategies Selection

We consider the problem of selecting the best variable-value strategy for solving a given problem in constraint programming. We show that the recent Embarrassingly Parallel Search method (EPS) can be
...

References

SHOWING 1-10 OF 25 REFERENCES

Algorithm Portfolio Design: Theory vs. Practice

TLDR
This work provides a detailed evaluation of the portfolio approach on distributions of hard combinatorial search problems and shows under what conditions the protfolio approach can have a dramatic computational advantage over the best traditional methods.

Boosting Combinatorial Search Through Randomization

TLDR
This work presents a general method for introducing controlled randomization into complete search algorithms and demonstrates speedups of several orders of magnitude for state-of-the-art complete search procedures running on hard, real-world problems.

An Economics Approach to Hard Computational Problems

TLDR
This method, based on notions of risk in economics, offers a computational portfolio design procedure that can be used for a wide range of problems, including the combinatorics of DNA sequencing and the completion of tasks in environments with resource contention, such as the World Wide Web.

Problem Structure in the Presence of Perturbations

TLDR
This work proposes a new benchmark domain that bridges the gap between the purely random instances and the highly structured problems, by introducing perturbations into a structured domain and demonstrates that the performance of search strategies designed to mimic direct constructive methods degrade surprisingly quickly in the presence of even minor perturbation.

"Go with the winners" algorithms

  • D. AldousU. Vazirani
  • Computer Science
    Proceedings 35th Annual Symposium on Foundations of Computer Science
  • 1994
TLDR
It is proved that the running time of the "go with the winners" scheme (to achieve 99% probability of success) is bounded by a polynomial in d and /spl kappa/.

Cooperative Strategies for Solving the Bicriteria Sparse Multiple Knapsack Problem

TLDR
This paper introduces the use of a cooperative problem solving team of heuristics that evolves algorithms for a given problem instance that uniformly improve solutions as compared to using predesigned heuristic algorithms.

Heavy-Tailed Distributions in Combinatorial Search

TLDR
It is shown how random restarts can effectively eliminate heavy-tailed behavior, thereby dramatically improving the overall performance of a search procedure.

Randomized Algorithms

TLDR
This book introduces the basic concepts in the design and analysis of randomized algorithms and presents basic tools such as probability theory and probabilistic analysis that are frequently used in algorithmic applications.

Optimal speedup of Las Vegas algorithms

TLDR
The authors describe a simple universal strategy S/sup univ/, with the property that, for any algorithm A, T(A,S/Sup univ/)=O (l/sub A/log(l/ sub A/)), which is the best performance that can be achieved, up to a constant factor, by any universal strategy.

On the Use of Integer Programming Models in AI Planning

TLDR
The main objective is to show that a carefully chosen IP formulation significantly improves the "strength" of the LP relaxation, and that the resultant LPs are useful in solving the IP and the associated planning problems.