Showing 1–2 of 2 results for author: Albers, K

Approximate Inference and Constrained Optimization

Authors: Tom Heskes, Kees Albers, Hilbert Kappen

Abstract: Loopy and generalized belief propagation are popular algorithms for approximate inference in Markov random fields and Bayesian networks. Fixed points of these algorithms correspond to extrema of the Bethe and Kikuchi free energy. However, belief propagation does not always converge, which explains the need for approaches that explicitly minimize the Kikuchi/Bethe free energy, such… ▽ More Loopy and generalized belief propagation are popular algorithms for approximate inference in Markov random fields and Bayesian networks. Fixed points of these algorithms correspond to extrema of the Bethe and Kikuchi free energy. However, belief propagation does not always converge, which explains the need for approaches that explicitly minimize the Kikuchi/Bethe free energy, such as CCCP and UPS. Here we describe a class of algorithms that solves this typically nonconvex constrained minimization of the Kikuchi free energy through a sequence of convex constrained minimizations of upper bounds on the Kikuchi free energy. Intuitively one would expect tighter bounds to lead to faster algorithms, which is indeed convincingly demonstrated in our simulations. Several ideas are applied to obtain tight convex bounds that yield dramatic speed-ups over CCCP. △ Less

Submitted 19 October, 2012; originally announced December 2012.

Comments: Appears in Proceedings of the Nineteenth Conference on Uncertainty in Artificial Intelligence (UAI2003)

Report number: UAI-P-2003-PG-313-320

Efficient Feasibility Analysis for Real-Time Systems with EDF Scheduling

Authors: Karsten Albers, Frank Slomka

Abstract: This paper presents new fast exact feasibility tests for uniprocessor real-time systems using preemptive EDF scheduling. Task sets which are accepted by previously described sufficient tests will be evaluated in nearly the same time as with the old tests by the new algorithms. Many task sets are not accepted by the earlier tests despite them beeing feasible. These task sets will be evaluated by… ▽ More This paper presents new fast exact feasibility tests for uniprocessor real-time systems using preemptive EDF scheduling. Task sets which are accepted by previously described sufficient tests will be evaluated in nearly the same time as with the old tests by the new algorithms. Many task sets are not accepted by the earlier tests despite them beeing feasible. These task sets will be evaluated by the new algorithms a lot faster than with known exact feasibility tests. Therefore it is possible to use them for many applications for which only sufficient test are suitable. Additionally this paper shows that the best previous known sufficient test, the best known feasibility bound and the best known approximation algorithm can be derived from these new tests. In result this leads to an integrated schedulability theory for EDF. △ Less

Submitted 25 October, 2007; originally announced October 2007.

Comments: Submitted on behalf of EDAA (http://www.edaa.com/)

Journal ref: Dans Design, Automation and Test in Europe - DATE'05, Munich : Allemagne (2005)