PASCAL - Pattern Analysis, Statistical Modelling and Computational Learning

Information Theory for Angular Data
Peter Harremoes
In: ISIT 2008, Canada(2008).


In this paper we study various aspect of information theory for data with values on a circle or on a sphere. We develop rate distortion theory with squared Euclidian distance as distortion function. It is shown that some ideas related to the minimum description length principle does not apply in this context. Various inequalities are proved. Pointwise convergence of a sequence of information projections is conjectured and some evidence is provided.

EPrint Type:Conference or Workshop Item (Paper)
Project Keyword:Project Keyword UNSPECIFIED
Subjects:Learning/Statistics & Optimisation
ID Code:3469
Deposited By:Peter Harremoes
Deposited On:11 February 2008