Title :
On Distributional Assumptions and Whitened Cosine Similarities
Author_Institution :
Fac. of Electr. Eng., Delft Univ. of Technol., Delft
fDate :
6/1/2008 12:00:00 AM
Abstract :
Recently, an interpretation of the whitened cosine similarity measure as a Bayes decision rule was proposed [1]. This communication makes the observation that some of the distributional assumptions made to derive this measure are very restrictive and, considered simultaneously even inconsistent.
Keywords :
Bayes methods; Bayes decision rule; distributional assumptions; whitened cosine similarities; Classifier design and evaluation; Statistical; Algorithms; Artificial Intelligence; Bayes Theorem; Computer Simulation; Image Enhancement; Image Interpretation, Computer-Assisted; Information Storage and Retrieval; Models, Statistical; Numerical Analysis, Computer-Assisted; Pattern Recognition, Automated;
Journal_Title :
Pattern Analysis and Machine Intelligence, IEEE Transactions on
DOI :
10.1109/TPAMI.2007.70838
