ABSTRACT
Learning Classifier System (LCS) is a class of rule-based learning algorithms, which combine reinforcement learning (RL) and genetic algorithm (GA) techniques to evolve a population of classifiers. The most prominent example is XCS, for which many variants have been proposed in the past, including XCSF for function approximation. Although XCSF is a promising candidate for supporting autonomy in computing systems, it still must undergo parameter optimization prior to deployment. However, in case the later deployment environment is unknown, a-priori parameter optimization is not possible, raising the need for XCSF to automatically determine suitable parameter values at run-time. One of the most important parameters is the error threshold, which can be interpreted as a target bound on the approximation error and has to be set according to the approximated function. To enable XCSF to reliably approximate functions unknown at design-time, we propose the use of an error threshold, which is adapted at run-time based on the currently achieved approximation error. Our experimental evaluation shows that the adaption mechanism automatically adjusts the error threshold to a suitable value. On six different target functions, XCSF with an adaptive error threshold leads to a worst-case approximation error that is 11.5% higher than the error achieved with the best-suited static threshold.
- Thomas Bäck. 1992. Self-Adaption in Genetic Algorithms. In Toward a practice of autonomous systems. Proceedings of the 1st European Conference on Artifical Life, Francisco J. Varela and Paul Bourgine (Eds.). MIT Press, 263--271.Google Scholar
- M.V. Butz, T. Kovacs, P.L. Lanzi, and S.W. Wilson. 2004. Toward a Theory of Generalization and Learning in XCS. IEEE Transactions on Evolutionary Computation 8, 1 (feb 2004), 28--46. Google ScholarDigital Library
- M.V. Butz, P.L. Lanzi, and S.W. Wilson. 2008. Function Approximation With XCS: Hyperellipsoidal Conditions, Recursive Least Squares, and Compaction. IEEE Transactions on Evolutionary Computation 12, 3 (jun 2008), 355--376. Google ScholarDigital Library
- Martin V. Butz, Pier Luca Lanzi, and Stewart W. Wilson. 2006. Hyper-Ellipsoidal Conditions in XCS: Rotation, Linear Approximation, and Solution Structure. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '06. ACM, New York, NY, USA, 1457--1464. Google ScholarDigital Library
- Martin V. Butz, Patrick Stalph, and Pier Luca Lanzi. 2008. Self-adaptive mutation in XCSF. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '08. ACM, New York, NY, USA, 1365--1372. Google ScholarDigital Library
- M. V. Butz and S. W. Wilson. 2002. An algorithmic description of XCS. Soft Computing - A Fusion of Foundations, Methodologies and Applications 6, 3-4 (jun 2002), 144--153. Google ScholarCross Ref
- Hai H. Dam, Chris Lokan, and Hussein A. Abbass. 2007. Evolutionary Online Data Mining: An Investigation in a Dynamic Environment. In Evolutionary Computation in Dynamic and Uncertain Environments, Shengxiang Yang, Yew-Soon Ong, and Yaochu Jin (Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 153--178. Google ScholarCross Ref
- Jeroen Eggermont and Jano I. van Hemert. 2001. Adaptive Genetic Programming Applied to New and Existing Simple Regression Problems. In Genetic Programming, Julian Miller, Marco Tomassini, Pier Luca Lanzi, Conor Ryan, Andrea G. B. Tettamanzi, and William B. Langdon (Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 23--35.Google ScholarDigital Library
- A.E. Eiben, R. Hinterding, and Z. Michalewicz. 1999. Parameter control in evolutionary algorithms. IEEE Transactions on Evolutionary Computation 3, 2 (jul 1999), 124--141. Google ScholarDigital Library
- Tim Hansmeier, Paul Kaufmann, and Marco Platzner. 2020. Enabling XCSF to Cope with Dynamic Environments via an Adaptive Error Threshold. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '20. ACM, New York, NY, USA. Google ScholarDigital Library
- Jürgen Hesser and Reinhard Männer. 1991. Towards an optimal mutation probability for genetic algorithms. In Parallel Problem Solving from Nature, Hans-Paul Schwefel and Reinhard Männer (Eds.). Springer Berlin Heidelberg, Berlin, Heidelberg, 23--32.Google Scholar
- Jacob Hurst and Larry Bull. 2001. Self-adaptation in classifier system controllers. Artificial Life and Robotics 5, 2 (jun 2001), 109--119. Google ScholarCross Ref
- Jacob Hurst and Larry Bull. 2002. A Self-Adaptive XCS. In Advances in Learning Classifier Systems - IWLCS 2001. Springer Berlin Heidelberg, 57--73. Google ScholarCross Ref
- J.O. Kephart and D.M. Chess. 2003. The vision of autonomic computing. Computer 36, 1 (jan 2003), 41--50. Google ScholarDigital Library
- Piere Luca Lanzi and Daniele Loiacono. 2009. XCSLib: The XCS Classifier System Library. Technical Report 2009005. Illinois Genetic Algorithms Laboratory, University of Illinois at Urbana-Champaign.Google Scholar
- Pier Luca Lanzi, Daniele Loiacono, Stewart W. Wilson, and David E. Goldberg. 2005. Extending XCSF beyond linear approximation. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '05. ACM, New York, NY, USA, 1827--1834. Google ScholarDigital Library
- Peter R. Lewis, Marco Platzner, Bernhard Rinner, Jim Tørresen, and Xin Yao (Eds.). 2016. Self-aware Computing Systems. Springer International Publishing. Google ScholarCross Ref
- H. B. Mann and D. R. Whitney. 1947. On a Test of Whether one of Two Random Variables is Stochastically Larger than the Other. The Annals of Mathematical Statistics 18, 1 (mar 1947), 50--60. Google ScholarCross Ref
- Christian Müller-Schloer, Hartmut Schmeck, and Theo Ungerer (Eds.). 2011. Organic Computing --- A Paradigm Shift for Complex Systems. Springer Basel. Google ScholarCross Ref
- Masaya Nakata, Will Browne, and Tomoki Hamagami. 2018. Theoretical adaptation of multiple rule-generation in XCS. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '18. ACM, New York, NY, USA, 482--489. Google ScholarDigital Library
- Masaya Nakata, Will Browne, Tomoki Hamagami, and Keiki Takadama. 2017. Theoretical XCS parameter settings of learning accurate classifiers. In Proceedings of the Genetic and Evolutionary Computation Conference - GECCO '17. ACM, New York, NY, USA, 473--480. Google ScholarDigital Library
- Miguel Nicolau, Alexandros Agapitos, Michael O'Neill, and Anthony Brabazon. 2015. Guidelines for defining benchmark problems in Genetic Programming. In IEEE Congress on Evolutionary Computation, CEC 2015 - Proceedings. IEEE, 1152--1159. Google ScholarCross Ref
- Anthony Stein, Christian Eymuller, Dominik Rauh, Sven Tomforde, and Jorg Hahner. 2016. Interpolation-based classifier generation in XCSF. In IEEE Congress on Evolutionary Computation, CEC 2016 - Proceedings. IEEE, 3990--3998. Google ScholarDigital Library
- Maciej Troć and Olgierd Unold. 2010. Self-adaptation of parameters in a learning classifier system ensemble machine. International Journal of Applied Mathematics and Computer Science 20, 1 (mar 2010), 157--174. Google ScholarCross Ref
- Stewart W. Wilson. 1995. Classifier Fitness Based on Accuracy. Evolutionary Computation 3, 2 (jun 1995), 149--175. Google ScholarDigital Library
- Stewart W. Wilson. 2002. Classifiers that approximate functions. Natural Computing 1, 2 (01 Jun 2002), 211--234. Google ScholarDigital Library
Index Terms
- An adaption mechanism for the error threshold of XCSF
Recommendations
Enabling XCSF to cope with dynamic environments via an adaptive error threshold
GECCO '20: Proceedings of the 2020 Genetic and Evolutionary Computation Conference CompanionThe learning classifier system XCSF is a variant of XCS employed for function approximation. Although XCSF is a promising candidate for deployment in autonomous systems, its parameter dependability imposes a significant hurdle, as a-priori parameter ...
An anticipatory approach to improve XCSF
GECCO '06: Proceedings of the 8th annual conference on Genetic and evolutionary computationXCSF is a novel version of learning classifier systems (LCS) which extends the typical concept of LCS by introducing computable classifier prediction. In XCSF Classifier prediction is computed as a linear combination of classifier inputs and a weight ...
Improving generalization in the XCSF classifier system using linear least-squares
GECCO '05: Proceedings of the 7th annual workshop on Genetic and evolutionary computationXCSF is an extension of XCS in which classifier prediction is computed as a linear combination of classifier inputs and a weight vector associated to each classifier. XCSF can adjust the weight vector of classifiers to evolve accurate piecewise linear ...
Comments