The Use of Computational Creativity Metrics to Evaluate Alternative Values for Clustering Algorithm Parameters
This paper presents some details on how to use concepts from computational creativity to inform the machine learning task of clustering. Specifically, clustering involves structuring exemplar-based knowledge. The novelty and usefulness of the way the knowledge ends up being structured can be measured. These are characteristics that traditionally computational creativity focuses on whereas machine learning doesn’t, but they can aid in selecting the best value for the parameters of the learning task. Doing so also provides us with a way to find an adequate balance between novelty and usefulness, something that still hasn’t been fully formalized in computational creativity. Thus both fields, machine learning and computational creativity, can benefit from this type of hybrid research.
KeywordsComputational creativity Clustering k-means
This work has been supported by Asociación Mexicana de Cultura, A.C. This work was done while the author was on sabbatical leave at the Computational Intelligence Laboratory (CIL), School of Computer Engineering, Nanyang Technological University, Singapore. The author gratefully acknowledges the opportunity enabled by Professor Erik Cambria to do research there. While at the CIL the author benefitted from discussions with Anupam Mondal and Ranjan Satapathy, who also provided and processed the medical domain data and produced the graphs shown in Figs. 2 and 3. Many thanks are also due to Daniel Espinosa Mireles de Villafranca at ITAM, who helped to polish the final version of this paper.
- 2.Mitchell, T.M.: Machine learning. McGraw-Hill, New York (1997)Google Scholar
- 3.Varshney, L.R., Pinel, F., Varshney, K.R., Schörgendorfer, A., Chee, Y.M.: Cognition as a part of computational creativity. In: Proceedings of the 12th IEEE International Conference on Cognitive Informatics and Cognitive Computing, pp. 36–43 (2013)Google Scholar
- 4.Ma, P., Zemmel, R.: Value of novelty? Nat. Rev. 1, 571–572 (2002)Google Scholar
- 5.Carlisle, P.R., Lakhani, K.R.: Innovation and the challenge of novelty: the novelty-confirmation-transformation cycle in software and science. Harvard Business School Working Paper 11-096 (2011)Google Scholar
- 8.Anbarasi, M.S., Mehata, K.M.: Enhanced k-means clustering for patient reported outcome. In: Proceedings of the Fourth WSEAS International Conference on Computer Engineering and Applications, pp. 19–26 (2010)Google Scholar
- 9.Pham, D.T., Dimov, S.S., Nguyen, C.D.: Selection of k in k-means clustering. J. Mech. Eng. Sci. 219(C), 103–119 (2004)Google Scholar
- 10.Mondal, A., Chaturvedi, I., Das, D., Bajpai, R., Bandyopadhyay, S.: Lexical resource for medical events: a polarity based approach. In: Workshop Proceedings of the Fifteenth IEEE International Conference on Data Mining (ICDM 2015), pp. 1302–1309 (2015)Google Scholar
- 11.Mondal, A., Das, D., Cambria, E., Bandyaopadhyay, S.: WME: sense, polarity, and affinity based concept resource for medical events. In: Proceedings of the Eighth Global WordNet Conference, pp. 242–246 (2016)Google Scholar
- 13.Grace, K., Maher, M.L.: Using computational creativity to guide data-intensive scientific discovery. In: Proceedings of the 2014 Association for the Advancement of Artificial Intelligence Workshop on Discovery Informatics, pp. 35–38. AAAI Press (2014)Google Scholar