Boosting as a Metaphor for Algorithm Design

Leyton‐Brown, Kevin; Nudelman, Eugene; Andrew, Galen; McFadden, Jim; Shoham, Yoav

doi:10.1007/978-3-540-45193-8_75

Cited by 38 publications

(29 citation statements)

References 4 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Starting from the observation that different solvers have complementary strengths and weaknesses and so they show different behaviours on different problem instances, the ideas of running multiple solvers in parallel or to select one solver based on the features of a given instance were introduced. These approaches have been named algorithm portfolios [19,20,21,7]. Portfolio research is an active research topic nowadays as shown in [22], with more than 150 papers and counting.…”

Section: The Algorithm Selection Problemmentioning

confidence: 99%

SNNAP: Solver-Based Nearest Neighbor for Algorithm Portfolios

Collautti

Malitsky

Mehta

et al. 2013

Advanced Information Systems Engineering

View full text Add to dashboard Cite

A Mamma e Papà e ai miei fratelli AbstractThe success of portfolio algorithms in competitions in the area of combinatorial problem solving, as well as in practice, has motivated interest in the development of new approaches to determine the best solver for the problem at hand. Yet, although there are a number of ways in which this decision can be made, it always relies on a rich set of features to identify and distinguish the structure of the problem instances.In this thesis, however, it is firstly shown how not all the features in the problem have the same relevancy. Then it is presented how one of the most successful portfolio approaches, ISAC, can be augmented by taking into account the past performance of solvers as part of the feature vector. Testing on a variety of SAT datasets, it is here proved how the new formulation continuously outperforms an unmodified/standard version of ISAC.This thesis presents a novel strategy that tackles the problem of algorithm portfolios applying machine learning techniques and improves a state-of-the-art portfolio algorithms approaches (ISAC).Starting from the basic assumptions behind the standard ISAC methodology that the given collection of features can be used to correctly identify the structure of each instance, an important step has been to be able to improve clustering, grouping together instances that prefer to be solved by the same solver. It has been created, and here is presented, a methodology to redefine the feature vector including past performance of the solvers. The final result is named SNNAP: Solver-based Nearest Neighbor for Algorithm Portfolios which is here presented.

show abstract

Section: The Algorithm Selection Problemmentioning

confidence: 99%

SNNAP: Solver-Based Nearest Neighbor for Algorithm Portfolios

Collautti

Malitsky

Mehta

et al. 2013

Advanced Information Systems Engineering

View full text Add to dashboard Cite

show abstract

“…The contributions in this thesis are related to a variety of existing research topics including: teamwork models [25,39,49,69], teamwork measures [7,26,35], algorithm selection [2,23,38,40], and dynamic coordination [17]. To understand the significance of the contribution of this thesis, we first present related work from these areas.…”

Section: Introductionmentioning

confidence: 99%

A Study of Dynamic Coordination Mechanisms

Rosenfeld¹

2009

The Knowledge Engineering Review

View full text Add to dashboard Cite

“…Intuitively, the purpose of 5 is to somehow select the "best" hi for solving or deciding x according to some criteria defined by algorithm designers. The design of selectors given an existing collection of heuristics in practice is called the algorithm selection problem [44] and has been studied in numerous contexts within artificial intelligence and operations research (see [37,36,32,17,25,10] for a sample). Our novelty, however, is to allow the use of a different performance measure for each heuristic.…”

Section: Introductionmentioning

confidence: 99%

Confronting hardness using a hybrid approach

Vassilevska¹,

Williams²,

Woo³

2006

Proceedings of the Seventeenth Annual ACM-SIAM Symposium on Discrete Algorithm - SODA '06

View full text Add to dashboard Cite

A hybrid algorithm is a collection of heuristics, paired with a polynomial time selector S that runs on the input to decide which heuristic should be executed to solve the problem. Hybrid algorithms are interesting in scenarios where the selector must decide between heuristics that are "good" with respect to different complexity measures. In this paper, we focus on hybrid algorithms with a "hardness-defying" property: for a problem II, there is a set of complexity measures {rrii} whereby II is known or conjectured to be hard (or unsolvable) for each ra^, but for each heuristic hi of the hybrid algorithm, one can give a complexity guarantee for hi on the instances of II that S selects for hi that is strictly better than rrii. For example, we show that for NP-hard problems such as MAX-E/c-LlN-p, LONGEST PATH and MINIMUM BANDWIDTH, a given instance can either be solved exactly in "sub-exponential" (2°^) time, or be approximated in polynomial time with an approximation ratio exceeding that of the known or conjectured inapproximability of the problem, assuming P 7^ NP. We also prove some inherent limitations to the design of hybrid algorithms that arise under the assumption that NP requires exponential time algorithms.

show abstract

Boosting as a Metaphor for Algorithm Design

Cited by 38 publications

References 4 publications

SNNAP: Solver-Based Nearest Neighbor for Algorithm Portfolios

SNNAP: Solver-Based Nearest Neighbor for Algorithm Portfolios

A Study of Dynamic Coordination Mechanisms

Confronting hardness using a hybrid approach

Contact Info

Product

Resources

About