E-mail: harremoes@ieee.org Mobile: +45 27 82 41 71 Skype: peterharremoes Temporary office: Lincoln Hall, Room 409A 1821 East-West Road Honolulu, Hawai'i 96848 USA Permanent address: |
Research newsHere is a list of references
related to my lecture series on Bregman divergences. I just finished a long paper with various bounds on tail probabilities in terms of the signed log-likelihood function arXiv: 1601.05179 . Although the inequalities are quite sharp I am sure even sharper inequalities of these types can be obtained. I think the results can be generalized to cover all exponential families with simple variance functions. Even more general inequalities may exist, but at the moment I don't know how to attack the general problem. Proper Scoring and Sufficiency In this paper I develop a general framework based on optimization that leads to Bregman divergences. With the extra condition of sufficiency we get information divergence. Now I am extending these results in two directions. I have just submitted the paper arXiv: 1601.07593 to ISIT 2016 where the sufficiency condition in applied to portfolio theory. The other direction is to single out the convex sets where a Bregman divergence that satisfies the sufficiency condition is unique. I hope to be able to state the condition in terms of the symmetries of the convex set. In the following paper arXiv:1601.04255 we obtain a lower bound on the rate of convergence in the law of small numbers, that seem to be asymptotically tight. Lattice theory of causation I an working on a
larger project where I want to see to what extend
concepts related to cause and effect can be studied
using lattice theory. This summer I presented some
derived results related to Lattices with
non-Shannon Inequalities. In January and
February I am visiting University of Hawai'i where
several lattice experts are situated. |
||
ResearchMy research is centred
on information theory. One of my interests is how to
use ideas from information theory to derive results
in probability theory. Many of the most important
reslts in probability theory are convergence
theorems, and many of these convergence theorems can
be reformulated so that they state that the entropy
of a system increases to a maximum or that a
divergence converge to a minimum. These ideas are
also relevant in the theory of statistical tests.
Recently I have formalized a method for deriving
Jeffreys prior as the optimal prior using the
minimum description lenght principle. Editorial workI have been Editor-in-Chief of the journal Entropy 2007-2011 and editor 2011-2015. This is a free online journal that publishes articles about entropy and related concepts, and their numerous applications. I serve as reviewer of numerous other
journals including: Mathematical Reviews, Entropy,
IEEE Trans. Inform. Theory, Physica A, JIPAM, and many
others. I just signed up to Publons
(Dec. 2015) where review activities will be
registrered. At the moment no reviews are registrered
at my profile, but I guess that will soon change.
EventsQuantum Lunch Talk at Math. Dept. Univ. Copenhagen 6/1 2016 I will give a talk entitled "Determinism and Causality. Which aspects are shared between classical physics and quantum physics?"I will be visiting University of Hawai'i Jan.-Feb. 2016. I will give a lecture series on Bregman divergences. The first will be held Thursday 21/1 in Holmes Hall. See flyer for detailed information. IGAIA IV I have been invited to give a talk at the conference Information Geometry and its Applications IV, to be held on June 1-17, 2016 at Liblice Castle, Czech Republic. The title of my talk will be announced later. ISIT 2016 10-15/7 2016, Barcelona, Spain. I have submitted some abstracts for this conference. Beyond IID 4 18-22/7 2016, Barcelona, Spain. This is a satellite conference to ISIT. LinksI have made a BibTeX file with a lot of items related to information theory. Information Theory
Society This is part of IEEE, which I am
member of. The page has a lot of useful links related to
information theory. SoftwareHere are links to some of my favorite software. Dropbox Very
convenient for storing and sharing documents. ColleguesHere are links to some persons that I use collaborate with. Andrew
Barron |