A probabilistic architecture for algorithm portfolios
Heuristic algorithms for logical reasoning are increasingly successful on computationally difficult problems such as satisfiability, and these solvers enable applications from circuit verification to software synthesis. Whether a problem instance can be solved, however, often depends in practice on...
Main Author: | |
---|---|
Format: | Others |
Language: | en_US |
Published: |
2013
|
Subjects: | |
Online Access: | http://hdl.handle.net/2152/19828 |
id |
ndltd-UTEXAS-oai-repositories.lib.utexas.edu-2152-19828 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-UTEXAS-oai-repositories.lib.utexas.edu-2152-198282015-09-20T17:14:11ZA probabilistic architecture for algorithm portfoliosSilverthorn, Bryan ConnorArtificial intelligenceMachine learningPropositional logicAlgorithm selectionAlgorithm portfoliosSatisfiabilityHeuristic algorithms for logical reasoning are increasingly successful on computationally difficult problems such as satisfiability, and these solvers enable applications from circuit verification to software synthesis. Whether a problem instance can be solved, however, often depends in practice on whether the correct solver was selected and its parameters appropriately set. Algorithm portfolios leverage past performance data to automatically select solvers likely to perform well on a given instance. Existing portfolio methods typically select only a single solver for each instance. This dissertation develops and evaluates a more general portfolio method, one that computes complete solver execution schedules, including repeated runs of nondeterministic algorithms, by explicitly incorporating probabilistic reasoning into its operation. This modular architecture for probabilistic portfolios (MAPP) includes novel solutions to three issues central to portfolio operation: first, it estimates solver performance distributions from limited data by constructing a generative model; second, it integrates domain-specific information by predicting instances on which solvers exhibit similar performance; and, third, it computes execution schedules using an efficient and effective dynamic programming approximation. In a series of empirical comparisons designed to replicate past solver competitions, MAPP outperforms the most prominent alternative portfolio methods. Its success validates a principled approach to portfolio operation, offers a tool for tackling difficult problems, and opens a path forward in algorithm portfolio design.text2013-04-05T14:50:01Z2012-052013-04-05May 20122013-04-05T14:50:01Zapplication/pdfhttp://hdl.handle.net/2152/19828en_US |
collection |
NDLTD |
language |
en_US |
format |
Others
|
sources |
NDLTD |
topic |
Artificial intelligence Machine learning Propositional logic Algorithm selection Algorithm portfolios Satisfiability |
spellingShingle |
Artificial intelligence Machine learning Propositional logic Algorithm selection Algorithm portfolios Satisfiability Silverthorn, Bryan Connor A probabilistic architecture for algorithm portfolios |
description |
Heuristic algorithms for logical reasoning are increasingly successful on computationally difficult problems such as satisfiability, and these solvers enable applications from circuit verification to software synthesis. Whether a problem instance can be solved, however, often depends in practice on whether the correct solver was selected and its parameters appropriately set. Algorithm portfolios leverage past performance data to automatically select solvers likely to perform well on a given instance. Existing portfolio methods typically select only a single solver for each instance. This dissertation develops and evaluates a more general portfolio method, one that computes complete solver execution schedules, including repeated runs of nondeterministic algorithms, by explicitly incorporating probabilistic reasoning into its operation. This modular architecture for probabilistic portfolios (MAPP) includes novel solutions to three issues central to portfolio operation: first, it estimates solver performance distributions from limited data by constructing a generative model; second, it integrates domain-specific information by predicting instances on which solvers exhibit similar performance; and, third, it computes execution schedules using an efficient and effective dynamic programming approximation. In a series of empirical comparisons designed to replicate past solver competitions, MAPP outperforms the most prominent alternative portfolio methods. Its success validates a principled approach to portfolio operation, offers a tool for tackling difficult problems, and opens a path forward in algorithm portfolio design. === text |
author |
Silverthorn, Bryan Connor |
author_facet |
Silverthorn, Bryan Connor |
author_sort |
Silverthorn, Bryan Connor |
title |
A probabilistic architecture for algorithm portfolios |
title_short |
A probabilistic architecture for algorithm portfolios |
title_full |
A probabilistic architecture for algorithm portfolios |
title_fullStr |
A probabilistic architecture for algorithm portfolios |
title_full_unstemmed |
A probabilistic architecture for algorithm portfolios |
title_sort |
probabilistic architecture for algorithm portfolios |
publishDate |
2013 |
url |
http://hdl.handle.net/2152/19828 |
work_keys_str_mv |
AT silverthornbryanconnor aprobabilisticarchitectureforalgorithmportfolios AT silverthornbryanconnor probabilisticarchitectureforalgorithmportfolios |
_version_ |
1716823059954925568 |