Abstract: Which features determine the beauty of a musical piece? What makes a melody a good one? The automatic classification of a melody or musical piece as "good" or "bad" resides on data-mining techniques and statistical tools, sometimes helped by some form of subjective evaluation. For automatic music composition many approaches have been used, but Genetic Algorithms, Evolutionary Programming and Particle Swarm Optimization showed to be the most promising ones, given their creative capabilities.
The critical issue in the above described tasks is music feature selection. Simple melodies (pre-processed in order to exclude non fundamental embellishments) or complete musical pieces can be analyzed in various perspectives such as Zipf's law [Manaris et al, 2003, 2007]. Classification tasks can then be supported by classifiers that may include a humanized perspective.
In fact these classifiers are a fundamental piece of the fitness functions of GA's, EP or PSO, responsible for the automatic creation of simple melodies, complete musical pieces or variations on a given theme. These algorithms can be used in stand-alone applications, sometimes complemented with local search techniques, as well as in a sequential way as proposed by [Reis C, Machado J, 2009].
However, aestetichal evaluation is strongly governed by cultural patterns. So, a brief introduction to Western and Eastern music is made and some formal principles that characterize and differentiate these styles are presented. Some of these principles can be easily incorporated into a fitness function. However, a complete and perfect fitness function must also take into account the capacity of emotion evocation as this has been the goal, for centuries, of most part of music styles [Reis C, Marques V, Machado, J, 2008]. Coincidence or not, Darwin's already mentioned the importance of emotional expression for survival and adaptation [Darwin C, 1872].

Brief Biography of the Speaker:
Viriato M. Marques was born at Lisbon, Portugal, in the year of 1957. He is a Coordinator Professor of DEIS - Computer Science Department - of ISEC - Engineering Institute of the Polytechnic Institute of Coimbra - and researcher at GECAD - Knowledge Engineering and Decision Support Group - of ISEP - Engineering Institute of Polytechnic Institute of Oporto. His research interests are Artificial Intelligence, Case-Based Reasoning, Data-Mining, Evolutionary Computation and Fuzzy Systems. He was President of the Scientific Board of ISEC, President of the Computer Science Department of ISEC, Coordinator of the IPC Committee for the European University Association evaluation at ISEC, co-founder of the Portuguese Chapter of IEEE Computational Intelligence Society and chair of the WSEAS conference FS'05, Lisbon, Portugal. He is the author of more than 60 papers published in the proceedings of international conferences, journals and books, such as: CBR for Diagnosis: Evidence Relevance and Case Adaptation, V.Marques, J.T.Farinha, A.Brito, 13th WSEAS Conference on Computers, Rhodes, Greece, 22-25 July 2009; Computational Intelligence Techniques in Engineering, V.Marques, C.Reis, L.Roseiro, J.T.Machado, 2nd Conference on Nonlinear Science and Complexity, Oporto, 28-31 July 2008 (to be published by Springer); Know how retention and divulgation with a Fuzzy CBR System, V.Marques, J.T.Farinha, A.Brito, N.Pincho, ISDA'07 - International Conference on Intelligent Systems Design and Applications, Rio de Janeiro, Brasil, 22-24 October 2007. He has been reviewer or member of the Scientific Committee of various International Conferences. Viriato M. Marques graduated in Electrical and Electronic Engineering by FCTUC - Faculty of Science and Technology of Coimbra University - and obtained his Ph.D. in Engineering Sciences at FEUP - Engineering Faculty of Oporto University. He has professional experience in the field of Information Systems as manager and technical responsible of some enterprises, and finished the piano course at Coimbra Music School.