@INPROCEEDINGS{KarydisRNI2010,
  AUTHOR =       {Ioannis Karydis and Milo\v{s} Radovanovi\'c and Alexandros Nanopoulos and Mirjana Ivanovi\'c},
  TITLE =        {Looking Through the 'Glass Ceiling': A Conceptual Framework for the Problems of Spectral Similarity},
  BOOKTITLE =    {International Society for Music Information Retrieval},
  YEAR =         {2010},
  pages =        {273--278},
  abstract =     {Spectral similarity measures have been shown to exhibit good performance in several Music Information Retrieval (MIR) applications. They are also known, however, to possess several undesirable properties, namely allowing the existence of hub songs (songs which frequently appear in nearest neighbor lists of other songs), “orphans” (songs which practically never appear), and difficulties in distinguishing the farthest from the nearest neighbor due to the concentration effect caused by high dimensionality of data space. In this paper we develop a conceptual framework that allows connecting all three undesired properties. We show that hubs and "orphans" are expected to appear in high-dimensional data spaces, and relate the cause of their appearance with the concentration property of distance / similarity measures. We verify our conclusions on realmusic data, examining groups of frames generated by Gaussian Mixture Models (GMMs), considering two similarity measures: Earth Mover’s Distance (EMD) in combination with Kullback-Leibler (KL) divergence, and Monte Carlo (MC) sampling. The proposed framework can be useful to MIR researchers to address problems of spectral similarity, understand their fundamental origins, and thus be able to develop more robust methods for their remedy.},
}