DocServer

On the Principle of Maximum Entropy on the Mean as a methodology for the regularization of inverse problems

Marechal, Pierre (1998) On the Principle of Maximum Entropy on the Mean as a methodology for the regularization of inverse problems. [Preprint]

[img]
Preview
Postscript
Download (178Kb) | Preview
    [img]
    Preview
    PDF
    Download (195Kb) | Preview

      Abstract

      In this paper, the bi-level structure of the Principle of Maximum Entropy on the Mean (PMEM) is examined. It is shown that the essential features of the PMEM can be derived in two different manners. One of these approaches relies on elementary results on nested optimization problems and suggests considerable generalizations of the principle. This sheds a new light on this type of inference mechanisms and, unexpectedly, stresses some conceptual weaknesses of the PMEM as a general methodology for solving linear inverse problems.

      Item Type: Preprint
      Additional Information: pubdom FALSE
      Subjects: UNSPECIFIED
      Faculty: UNSPECIFIED
      Depositing User: Users 1 not found.
      Date Deposited: 25 Nov 2003
      Last Modified: 21 Apr 2010 11:13
      URI: https://docserver.carma.newcastle.edu.au/id/eprint/217

      Actions (login required)

      View Item