neural networks research group
Algorithm portfolio methods operate in problem domains for which there are multiple algorithms with complementary strengths. The portfolio method applies patterns learned from experience to better allocate computational resources among the algorithms, attempting to apply each algorithm primarily to those problem instances to which it is best suited. Applications of the methods developed in this work include SAT and answer set programming.
Surviving Solver Sensitivity: An ASP Practitioner's Guide
Bryan Silverthorn, Yuliya Lierler and Marius Schneider
International Conference on Logic Programming (ICLP)
A Probabilistic Architecture for Algorithm Portfolios
PhD Thesis, Department of Computer Science, The University of Texas at Austin, 2012.
Latent Class Models for Algorithm Portfolio Methods
Bryan Silverthorn and Risto Miikkulainen
Proceedings of the Twenty-Fourth AAAI Conference on Artificial Intelligence
bsilvert [at] cs utexas edu
Borg: A General-Purpose Algorithm Portfolio System
2009 - 2013
Model-Based Visualization of Solver Performance Data
The borg project
includes a practical algorithm...