%0 Journal Article %1 curry:2007:SMC %A Curry, Robert %A Lichodzijewski, Peter %A Heywood, Malcolm I. %D 2007 %J IEEE Transactions on Systems, Man, and Cybernetics: Part B - Cybernetics %K DSS, RSS, active algorithms, casGP classification, data, genetic hierarchical learning, linear programming, unbalanced %N 4 %P 1065--1073 %R doi:10.1109/TSMCB.2007.896406 %T Scaling Genetic Programming to Large Datasets Using Hierarchical Dynamic Subset Selection %U http://www.cs.dal.ca/~mheywood/X-files/GradPubs.html#curry %V 37 %X The computational overhead of Genetic Programming (GP) may be directly addressed without recourse to hardware solutions using active learning algorithms based on the Random or Dynamic Subset Selection heuristics (RSS or DSS). This work begins by presenting a family of hierarchical DSS algorithms: RSS-DSS, cascaded RSS-DSS, and the Balanced Block DSS algorithm; where the latter has not been previously introduced. Extensive benchmarking over four unbalanced real-world binary classification problems with 30,000 to 500,000 training exemplars demonstrates that both the cascade and Balanced Block algorithms are able to reduce the likelihood of degenerates, whilst providing a significant improvement in classification accuracy relative to the original RSS-DSS algorithm. Moreover, comparison with GP trained without an active learning algorithm indicates that classification performance is not compromised, while training is completed in minutes as opposed to half a day.

@article{curry:2007:SMC, abstract = {The computational overhead of Genetic Programming (GP) may be directly addressed without recourse to hardware solutions using active learning algorithms based on the Random or Dynamic Subset Selection heuristics (RSS or DSS). This work begins by presenting a family of hierarchical DSS algorithms: RSS-DSS, cascaded RSS-DSS, and the Balanced Block DSS algorithm; where the latter has not been previously introduced. Extensive benchmarking over four unbalanced real-world binary classification problems with 30,000 to 500,000 training exemplars demonstrates that both the cascade and Balanced Block algorithms are able to reduce the likelihood of degenerates, whilst providing a significant improvement in classification accuracy relative to the original RSS-DSS algorithm. Moreover, comparison with GP trained without an active learning algorithm indicates that classification performance is not compromised, while training is completed in minutes as opposed to half a day.}, added-at = {2008-06-19T17:35:00.000+0200}, author = {Curry, Robert and Lichodzijewski, Peter and Heywood, Malcolm I.}, biburl = {https://www.bibsonomy.org/bibtex/258e431152c0a65d21837f62e9c178342/brazovayeye}, doi = {doi:10.1109/TSMCB.2007.896406}, email = {mheywood@cs.dal.ca}, interhash = {1347dfae3c07a84d2274daa07f299376}, intrahash = {58e431152c0a65d21837f62e9c178342}, issn = {1083-4419}, journal = {IEEE Transactions on Systems, Man, and Cybernetics: Part B - Cybernetics}, keywords = {DSS, RSS, active algorithms, casGP classification, data, genetic hierarchical learning, linear programming, unbalanced}, month = {August}, notes = {max prog length=8, comparsion with lilGP, binary classification, unbalanced training sets, selecting balanced training subsets, page based crossover}, number = 4, pages = {1065--1073}, size = {9 pages}, timestamp = {2008-06-19T17:38:17.000+0200}, title = {Scaling Genetic Programming to Large Datasets Using Hierarchical Dynamic Subset Selection}, url = {http://www.cs.dal.ca/~mheywood/X-files/GradPubs.html#curry}, volume = 37, year = 2007 }