eVidenZ

eVidenZ is a Dempster-Shafer (D-S) engine developed by the LRDE (Epita Research and Development laboratory). This free C++ implementation is currently a research prototype, but it already provides all the structures and basic algorithms for designing efficient D-S engines.

Dempster-Shafer Theory

Dempster-Shafer theory, also known as theory of evidence or belief functions theory, is one of the most promising alternatives to the common probabilistics frameworks used in the field of quantitative reasoning and decision under uncertainty. This theory has been introduced in the late seventies by Glenn Shafer as a way of representing epistemic or uncertain knowledge, starting from the work of Arthur Dempster, Shafer's advisor. One of the particularities of this framework is its ability to represent total ignorance and to reason with partial ignorance by fusioning incomplete pieces of information.

Related Projects

• Adhoc

Availability

eVidenZ is free software. It is developed under the conditions of the GNU General Public License version 2.

First prototype will be released soon.

Documentation

You can find here our BibTeX bibliography containing interesting articles about different aspects of Dempster-Shafer theory.

We highlight here some references that helped us understanding both theoretical and practical aspects of D-S theory:

• theoretical pointers:
  • dempster67: a fondamental paper from Dempster, helpful to understand D-S historical bases. Some aspects are tediously described, and the explanations given are not the most "update-to-date" ones. (i.e. the probabilistic base described here is not actually the most acceptable one).
  • sh76a: another fondamental paper, from Shafer, Dempster's disciple. Shafer extends and correct Dempster's explanations about D-S theory.
  • sm94b: clear and well-argumented, this paper gives Smets' point of view on D-S theoretical basement. Unavoidable to understand clearly that belief masses are not probabilities. Beware that the vision given in this paper is Smets' Transferable Belief Model (TBM) one. Some papers won't agree with it. Anyway, it is the most reliable vision to us.

• implementation pointers:
  • lehmann01: a very helpful paper that gathers the description of fundamental algorithms. eVidenZ' core is mainly based on it.
  • xuke94: an advanced paper on D-S computations: Moebius transform and hyper-trees are introduced

• books:
  • ykf94: a collection of articles about D-S theory. From theoretical to practical aspects, this book covers a large range of D-S-related issues.

TODO List

• study the feasibility/relevance of "delayed valuation" (cf. technical report below)
• study data structures smarter than STL maps (i.e., replace them with hash tables)
• study the possibility of generating code through external tools

Publications

Nicolas Burrus and David Lesage. Theory of Evidence. CSI Seminar 0307 September 2003

Nicolas Burrus. Evidence Theory (part 1): Theoretical aspects. CSI Seminar

David Lesage. Evidence Theory (part 2): Implementation Issues and Applications. CSI Seminar
to top