Details
| Original language | English |
|---|---|
| Title of host publication | KI 2020 |
| Subtitle of host publication | Advances in Artificial Intelligence - 43rd German Conference on AI, Proceedings |
| Editors | Ute Schmid, Diedrich Wolter, Franziska Klügl |
| Publisher | Springer Science and Business Media Deutschland GmbH |
| Pages | 59-72 |
| Number of pages | 14 |
| ISBN (print) | 9783030582845 |
| Publication status | Published - 2020 |
| Externally published | Yes |
| Event | 43rd German Conference on Artificial Intelligence, KI 2020 - Bamberg, Germany Duration: 21 Sept 2020 → 25 Sept 2020 |
Publication series
| Name | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
|---|---|
| Volume | 12325 LNAI |
| ISSN (Print) | 0302-9743 |
| ISSN (electronic) | 1611-3349 |
Abstract
Algorithm selection (AS) is defined as the task of automatically selecting the most suitable algorithm from a set of candidate algorithms for a specific instance of an algorithmic problem class. While suitability may refer to different criteria, runtime is of specific practical relevance. Leveraging empirical runtime information as training data, the AS problem is commonly tackled by fitting a regression function, which can then be used to estimate the candidate algorithms’ runtimes for new problem instances. In this paper, we develop a new approach to algorithm selection that combines regression with ranking, also known as learning to rank, a problem that has recently been studied in the realm of preference learning. Since only the ranking of the algorithms is eventually needed for the purpose of selection, the precise numerical estimation of runtimes appears to be a dispensable and unnecessarily difficult problem. However, discarding the numerical runtime information completely seems to be a bad idea, as we hide potentially useful information about the algorithms’ performance margins from the learner. Extensive experimental studies confirm the potential of our hybrid approach, showing that it often performs better than pure regression and pure ranking methods.
Keywords
- Algorithm selection, Combined ranking and regression, Hybrid loss optimization
ASJC Scopus subject areas
- Mathematics(all)
- Theoretical Computer Science
- Computer Science(all)
- General Computer Science
Cite this
- Standard
- Harvard
- Apa
- Vancouver
- BibTeX
- RIS
KI 2020: Advances in Artificial Intelligence - 43rd German Conference on AI, Proceedings. ed. / Ute Schmid; Diedrich Wolter; Franziska Klügl. Springer Science and Business Media Deutschland GmbH, 2020. p. 59-72 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 12325 LNAI).
Research output: Chapter in book/report/conference proceeding › Conference contribution › Research › peer review
}
TY - GEN
T1 - Hybrid ranking and regression for algorithm selection
AU - Hanselle, Jonas
AU - Tornede, Alexander
AU - Wever, Marcel
AU - Hüllermeier, Eyke
N1 - Publisher Copyright: © Springer Nature Switzerland AG 2020.
PY - 2020
Y1 - 2020
N2 - Algorithm selection (AS) is defined as the task of automatically selecting the most suitable algorithm from a set of candidate algorithms for a specific instance of an algorithmic problem class. While suitability may refer to different criteria, runtime is of specific practical relevance. Leveraging empirical runtime information as training data, the AS problem is commonly tackled by fitting a regression function, which can then be used to estimate the candidate algorithms’ runtimes for new problem instances. In this paper, we develop a new approach to algorithm selection that combines regression with ranking, also known as learning to rank, a problem that has recently been studied in the realm of preference learning. Since only the ranking of the algorithms is eventually needed for the purpose of selection, the precise numerical estimation of runtimes appears to be a dispensable and unnecessarily difficult problem. However, discarding the numerical runtime information completely seems to be a bad idea, as we hide potentially useful information about the algorithms’ performance margins from the learner. Extensive experimental studies confirm the potential of our hybrid approach, showing that it often performs better than pure regression and pure ranking methods.
AB - Algorithm selection (AS) is defined as the task of automatically selecting the most suitable algorithm from a set of candidate algorithms for a specific instance of an algorithmic problem class. While suitability may refer to different criteria, runtime is of specific practical relevance. Leveraging empirical runtime information as training data, the AS problem is commonly tackled by fitting a regression function, which can then be used to estimate the candidate algorithms’ runtimes for new problem instances. In this paper, we develop a new approach to algorithm selection that combines regression with ranking, also known as learning to rank, a problem that has recently been studied in the realm of preference learning. Since only the ranking of the algorithms is eventually needed for the purpose of selection, the precise numerical estimation of runtimes appears to be a dispensable and unnecessarily difficult problem. However, discarding the numerical runtime information completely seems to be a bad idea, as we hide potentially useful information about the algorithms’ performance margins from the learner. Extensive experimental studies confirm the potential of our hybrid approach, showing that it often performs better than pure regression and pure ranking methods.
KW - Algorithm selection
KW - Combined ranking and regression
KW - Hybrid loss optimization
UR - http://www.scopus.com/inward/record.url?scp=85091138296&partnerID=8YFLogxK
U2 - 10.1007/978-3-030-58285-2_5
DO - 10.1007/978-3-030-58285-2_5
M3 - Conference contribution
SN - 9783030582845
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 59
EP - 72
BT - KI 2020
A2 - Schmid, Ute
A2 - Wolter, Diedrich
A2 - Klügl, Franziska
PB - Springer Science and Business Media Deutschland GmbH
T2 - 43rd German Conference on Artificial Intelligence, KI 2020
Y2 - 21 September 2020 through 25 September 2020
ER -