Pool-Based Realtime Algorithm Configuration: A Preselection Bandit Approach

Mesaoudi-Paul, Adil El; Bengs, Viktor; Hüllermeier, Eyke; Tierney, Kevin

doi:10.1007/978-3-030-53552-0_22

“…In offline configuration, the ranking can be used to generate new configurations (part of the next challenge), whereas in realtime configuration the rankings can be critical to deciding which configurations are allowed to try to solve the current instance. Methods for ranking and comparing include using empirical hardness models Leyton-Brown et al (2009) or more general surrogates as in GGA++ (Ansótegui et al, 2015) or SMAC , statistics (López-Ibánez et al, 2016), the TrueSkill mechanism (Fitzgerald et al, 2015), bandits (El Mesaoudi-Paul et al, 2020b). Nonetheless, there is undoubtedly still room for improvement, using perhaps new preference learning techniques (as have been used for AS in Hanselle et al (2020)) or deep learning models.…”

Section: Novel Ac Methodologiesmentioning

confidence: 99%

A Survey of Methods for Automated Algorithm Configuration

Schede¹,

Brandt²,

Tornede³

et al. 2022

Preprint

Self Cite

View full text Add to dashboard Cite

Algorithm configuration (AC) is concerned with the automated search of the most suitable parameter configuration of a parametrized algorithm. There is currently a wide variety of AC problem variants and methods proposed in the literature. Existing reviews do not take into account all derivatives of the AC problem, nor do they offer a complete classification scheme. To this end, we introduce taxonomies to describe the AC problem and features of configuration methods, respectively. We review existing AC literature within the lens of our taxonomies, outline relevant design choices of configuration approaches, contrast methods and problem variants against each other, and describe the state of AC in industry. Finally, our review provides researchers and practitioners with a look at future research directions in the field of AC.

show abstract

“…In offline configuration, the ranking can be used to generate new configurations (part of the next challenge), whereas in realtime configuration the rankings can be critical to deciding which configurations are allowed to try to solve the current instance. Methods for ranking and comparing include using empirical hardness models Leyton-Brown et al (2009) or more general surrogates as in GGA++ (Ansótegui et al, 2015) or SMAC ), statistics (López-Ibánez et al, 2016, the TrueSkill mechanism (Fitzgerald et al, 2015), bandits (El Mesaoudi-Paul et al, 2020b). Nonetheless, there is undoubtedly still room for improvement, using perhaps new preference learning techniques (as have been used for AS in Hanselle et al (2020)) or deep learning models.…”

Section: Novel Ac Methodologiesmentioning

confidence: 99%

A Survey of Methods for Automated Algorithm Configuration

Schede

¹

,

Brandt

²

,

Tornede

³

et al. 2022

jair

Self Cite

View full text Add to dashboard Cite

Algorithm configuration (AC) is concerned with the automated search of the most suitable parameter configuration of a parametrized algorithm. There is currently a wide variety of AC problem variants and methods proposed in the literature. Existing reviews do not take into account all derivatives of the AC problem, nor do they offer a complete classification scheme. To this end, we introduce taxonomies to describe the AC problem and features of configuration methods, respectively. We review existing AC literature within the lens of our taxonomies, outline relevant design choices of configuration approaches, contrast methods and problem variants against each other, and describe the state of AC in industry. Finally, our review provides researchers and practitioners with a look at future research directions in the field of AC.

show abstract

“…Another related branch of the literature is realtime algorithm configuration (Fitzgerald et al 2014;Fitzgerald, Malitsky, and O'Sullivan 2015;El Mesaoudi-Paul et al 2020), where in contrast to our setting, one seeks to find a suitable configuration of one single target algorithm (instead of the algorithm itself) for incoming problem instances.…”

Section: Related Workmentioning

confidence: 99%

Machine Learning for Online Algorithm Selection under Censored Feedback

Tornede

¹

,

Bengs

²

,

Hüllermeier

³

2022

AAAI

Self Cite

View full text Add to dashboard Cite

In online algorithm selection (OAS), instances of an algorithmic problem class are presented to an agent one after another, and the agent has to quickly select a presumably best algorithm from a fixed set of candidate algorithms. For decision problems such as satisfiability (SAT), quality typically refers to the algorithm's runtime. As the latter is known to exhibit a heavy-tail distribution, an algorithm is normally stopped when exceeding a predefined upper time limit. As a consequence, machine learning methods used to optimize an algorithm selection strategy in a data-driven manner need to deal with right-censored samples, a problem that has received little attention in the literature so far. In this work, we revisit multi-armed bandit algorithms for OAS and discuss their capability of dealing with the problem. Moreover, we adapt them towards runtime-oriented losses, allowing for partially censored data while keeping a space- and time-complexity independent of the time horizon. In an extensive experimental evaluation on an adapted version of the ASlib benchmark, we demonstrate that theoretically well-founded methods based on Thompson sampling perform specifically strong and improve in comparison to existing methods.

show abstract

Pool-Based Realtime Algorithm Configuration: A Preselection Bandit Approach

Cited by 7 publications

References 18 publications

A Survey of Methods for Automated Algorithm Configuration

A Survey of Methods for Automated Algorithm Configuration

A Survey of Methods for Automated Algorithm Configuration

Machine Learning for Online Algorithm Selection under Censored Feedback

Contact Info

Product

Resources

About