Research

The agenda for the Ninth International Workshop on Learning Classifier Systems (IWLCS’2006) can be found here. The workshop is coming with a list of very interesting papers and topics. We are looking forward to another edition of the workshop crowded of new and exiting ideas. If you are in GECCO, do not let it pass by ;). By the way, if you want to dig a little further into the learning classifier systems (LCS) and other genetics-based machine learning (GBML) world, check the website of the NCSA/IlliGAL Gathering on Evolutionary Learning (NIGEL’2006). It contains the slides and videos of the talks on cutting-edge LCS and GBML research. ...

Yes, with the new update of the website GALE was unreachable for a while. Here it is the original code again. Please, take a few minutes to read this post. Hope you find it useful. Enjoy :D GALE? GALE (Genetic and Artificial Life Environment) is fine-grained parallel genetic algorithm for data mining. Its main contributions are simplicity and its knowledge-independent model. The simplicity of GALE relies in its fine-grained parallelism based on spreading the population (feasible solutions to the classification task) over a 2D grid. Thus, artificial evolution can be easily modeled in terms of neighborhood relations. These neighborhood relations define GALE as a massive parallel evolutionary model. On the other hand, GALE does not constrain the knowledge representation. It can evolve indistinctly rules, instances, partially defined instances, and decision trees (orthogonal, oblique, and multivariate based on nearest neighbor). ...

The list of papers to be presented at the Ninth International Workshop on Learning Classifier Systems (IWLCS 2006) can be found here.

by Xavier Llorà (2006). IlliGAL TR No 2006017. Link to the PDF. Link to the Java code Abstract This technical report describes how to compute the fitness of a rule for an arbitrary size multiplexer without doing any instance matching. Pittsburgh-style learning classifier systems require the accuracy and the error of a rule to compute a fitness that promotes maximally accurate and maximally general rules. The accuracy (α) may be computed as the proportion of overall examples correctly classified, and the error (ε) is the proportion of incorrect classifications issued. Once the accuracy and error of a rule are known, the fitness can be computed as f(r)=α(r)*ε(r). This technical note shows how to computed the fitness only by inspecting the rule, requiring a time proportional to number of possible address values O(2^|a|) instead of the O(2^l) that requires a traditional rule matching strategy against all the possible instances. The proposed method makes tractable for Pittsburgh learning classifier systems multiplexer problems larger than the 11-input one. ...

by Xavier Llorà and David E. Goldberg, Yukio Ohsawa, Naohiro Matsumura, Yuichi Washida, Hiroshi Tamura, Masataka Yoshikawa, Michael Welge, Loretta Auvil, Duane Searshmith, Kei Ohnishi, and Chen-Ju Chao (2006). New Mathematics and Natural Computation, World Scientific, pp. 2(1):85–100. Link to the Journal publication. Abstract Creativity protocols and methodologies tend to be time consuming if applied manually. This paper presents how information technologies can support innovation and creativity for collaborative scenario creation and discussion. The fusion of change discovery, genetics algorithms, and computer-supported collaborative tools provide computational models of innovation and creativity. The proposed technology allows groups of participants in a creative processes to have pervasive access to the analysis of the current scenario in real time. This paper introduces such innovation technologies gathered in the DISCUS project, and summarizes initial successful usages of DISCUS on marketing research workshops. ...