Title :
Feature Filtering for Instance-Specific Algorithm Configuration
Author :
Kroer, Christian ; Malitsky, Yuri
Author_Institution :
IT-Univ. of Copenhagen, Copenhagen, Denmark
Abstract :
Instance-Specific Algorithm Configuration (ISAC) is a novel general technique for automatically generating and tuning algorithm portfolios. The approach has been very successful in practice, but up to now it has been committed to using all the features it was provided. However, traditional feature filtering techniques are not applicable, requiring multiple computationally expensive tuning steps during the evaluation stage. To this end, we show three new evaluation functions that use precomputed runtimes of a collection of untuned solvers to quickly evaluate subsets of features. One of our proposed functions even shows how to generate such an effective collection of solvers when only one highly parameterized solver is available. Using these new functions, we show that the number of features used by ISAC can be reduced to less than a quarter of the original number while often providing significant performance gains. We present numerical results on both SAT and CP domains.
Keywords :
computability; constraint handling; feature extraction; learning (artificial intelligence); set theory; ISAC; SAT; constraint programming; evaluation function; feature filtering; feature subset; instance-specific algorithm configuration; tuning algorithm portfolio; Benchmark testing; Clustering algorithms; Portfolios; Prediction algorithms; Runtime; Training; Tuning; CP; SAT; algorithm configuration; feature selection;
Conference_Titel :
Tools with Artificial Intelligence (ICTAI), 2011 23rd IEEE International Conference on
Conference_Location :
Boca Raton, FL
Print_ISBN :
978-1-4577-2068-0
Electronic_ISBN :
1082-3409
DOI :
10.1109/ICTAI.2011.132