on two composition operator in dempster shafer theory
play

On Two Composition Operator in Dempster-Shafer Theory Radim Jirou - PowerPoint PPT Presentation

On Two Composition Operator in Dempster-Shafer Theory Radim Jirou sek Faculty of Management, Jind rich uv Hradec, Czech Republic & Institute of Information Theory and Automation Czech Academy of Sciences, Prague ISIPTA 2015


  1. On Two Composition Operator in Dempster-Shafer Theory Radim Jirouˇ sek Faculty of Management, Jindˇ rich˚ uv Hradec, Czech Republic & Institute of Information Theory and Automation Czech Academy of Sciences, Prague ISIPTA 2015 Pescara, Italy July 20 – 24, 2015

  2. Outline of the Lecture What is a composition?

  3. Outline of the Lecture What is a composition? Operators of Composition in Dempster-Shafer theory

  4. Outline of the Lecture What is a composition? Operators of Composition in Dempster-Shafer theory Properties of the Operators of Composition

  5. Outline of the Lecture What is a composition? Operators of Composition in Dempster-Shafer theory Properties of the Operators of Composition Conclusions - An Open Problem

  6. What is a composition? Composition assembles knowledge in the framework of uncertainty calculus. ◮ In probability theory: probability distributions. ◮ In possibility theory: possibility distributions. ◮ In Dempster-Shafer theory: basic probability assignments (commonality functions). ◮ In Valuation-Based systems: valuations.

  7. Required properties of compositions ◮ K and L are sets of variables; Let ◮ κ is a valuation for K , λ for L : κ ( K ), λ ( L ); Required properties of the composition: • κ ⊲ λ is a valuation for K ∪ L ;

  8. Required properties of compositions ◮ K and L are sets of variables; Let ◮ κ is a valuation for K , λ for L : κ ( K ), λ ( L ); Required properties of the composition: • κ ⊲ λ is a valuation for K ∪ L ; • If κ and λ are consistent (i.e., κ ↓ K ∩ L = λ ↓ K ∩ L ), then κ ⊲ λ = λ ⊲ κ is their joint extension.

  9. Required properties of compositions ◮ K and L are sets of variables; Let ◮ κ is a valuation for K , λ for L : κ ( K ), λ ( L ); Required properties of the composition: • κ ⊲ λ is a valuation for K ∪ L ; • If κ and λ are consistent (i.e., κ ↓ K ∩ L = λ ↓ K ∩ L ), then κ ⊲ λ = λ ⊲ κ is their joint extension. • If κ and λ are not consistent, then κ ⊲ λ is an extension of κ (as similar to κ as possible).

  10. Required properties of compositions ◮ K and L are sets of variables; Let ◮ κ is a valuation for K , λ for L : κ ( K ), λ ( L ); Required properties of the composition: • κ ⊲ λ is a valuation for K ∪ L ; • If κ and λ are consistent (i.e., κ ↓ K ∩ L = λ ↓ K ∩ L ), then κ ⊲ λ = λ ⊲ κ is their joint extension. • If κ and λ are not consistent, then κ ⊲ λ is an extension of κ (as similar to κ as possible). • If L ⊆ M ⊆ K ∪ L , then ( κ ⊲ λ ) ↓ M = κ ↓ K ∩ M .

  11. Required properties of compositions ◮ K and L are sets of variables; Let ◮ κ is a valuation for K , λ for L : κ ( K ), λ ( L ); Required properties of the composition: • κ ⊲ λ is a valuation for K ∪ L ; • If κ and λ are consistent (i.e., κ ↓ K ∩ L = λ ↓ K ∩ L ), then κ ⊲ λ = λ ⊲ κ is their joint extension. • If κ and λ are not consistent, then κ ⊲ λ is an extension of κ (as similar to κ as possible). • If L ⊆ M ⊆ K ∪ L , then ( κ ⊲ λ ) ↓ M = κ ↓ K ∩ M . • . . . (e.g. associativity under special conditions)

  12. Required properties of compositions But, keep in mind that • composition is idempotent;

  13. Required properties of compositions But, keep in mind that • composition is idempotent; • composition is not used for knowledge updating;

  14. Required properties of compositions But, keep in mind that • composition is idempotent; • composition is not used for knowledge updating; • composition is different from combination,

  15. Required properties of compositions But, keep in mind that • composition is idempotent; • composition is not used for knowledge updating; • composition is different from combination, but it can be defined by combination ⊕ (and its reversal) κ ⊲ λ = κ ⊕ λ ⊖ λ ↓ K ∩ L , which prevents from the double-counting of knowledge.

  16. Dempster’s Operator of Composition R. Jirouˇ sek and P.P. Shenoy. Compositional models in valuation-based systems. IJAR, 2014(55), 1, 277–293. Definition. For commonality functions θ 1 on X K and θ 2 on X L ( K � = ∅ � = L ) the commonality function of their composition θ 1 d ⊲ θ 2 is defined for each nonempty c ⊆ X K ∪ L by the formula: α − 1 θ 1 ( c ↓ K ) · θ 2 ( c ↓ L )  if θ ↓ K ∩ L ( c ↓ K ∩ L ) > 0 ,  2 ( θ 1 d ⊲ θ 2 )( c ) = θ ↓ K ∩ L ( c ↓ K ∩ L ) 2 0 otherwise,  where α is a normalization constant defined as θ 1 ( d ↓ K ) · θ 2 ( d ↓ L ) � ( − 1) | d | +1 α = . θ ↓ K ∩ L ( d ↓ K ∩ L ) 2 d ∈ 2 X K ∪ L : θ ↓ K ∩ L ( d ↓ K ∩ L ) > 0 2

  17. Factorizing Operator of Composition R. Jirouˇ sek, J. Vejnarov´ a and M. Daniel. Compositional models of belief functions. ISIPTA 2007, 243–252. Definition. For basic assignments µ 1 on X K and µ 2 on X L ( K � = ∅ � = L ) their composition µ 1 f ⊲ µ 2 is defined for each nonempty c ⊆ X K ∪ L by one of the following formulae: ⊳ c ↓ L then (i) if µ ↓ K ∩ L ( c ↓ K ∩ L ) > 0 and c = c ↓ K ⊲ 2 ( µ 1 f ⊲ µ 2 )( c ) = µ 1 ( c ↓ K ) · µ 2 ( c ↓ L ) ; µ ↓ K ∩ L ( c ↓ K ∩ L ) 2 ( c ↓ K ∩ L ) = 0 and c = c ↓ K × X L \ K then (ii) if µ ↓ K ∩ L 2 ( µ 1 f ⊲ µ 2 )( c ) = m 1 ( c ↓ K ); (iii) in all other cases, ( µ 1 f ⊲ µ 2 )( c ) = 0 .

  18. Properties of the Operators of Composition ◮ Both operators meet all the properties of composition: • can be used for multidimensional model representation; • makes local computations in decomposable models possible.

  19. Properties of the Operators of Composition ◮ Both operators meet all the properties of composition: • can be used for multidimensional model representation; • makes local computations in decomposable models possible. ◮ Both operators yield the same result if applied • to basic assignments defined for disjoint sets of variables; • to Bayesian basic assignments.

  20. Properties of the Operators of Composition ◮ Both operators meet all the properties of composition: • can be used for multidimensional model representation; • makes local computations in decomposable models possible. ◮ Both operators yield the same result if applied • to basic assignments defined for disjoint sets of variables; • to Bayesian basic assignments. ◮ Both operators can be used to solve a marginal problem by the application of IPFP; for both of them it holds that if the process converges then the result is a joint extension of all the input basic assignments.

  21. Properties of the Operators of Composition Differences ◮ IPFP: for factorizing operator Csisz´ ar’s convergence theorem holds true, it does not hold for Dempster operator ( R. Jirouˇ sek, V. Kratochv´ ıl. On Open Problems Connected with Application of the Iterative Proportional Fitting Procedure to Belief Functions. ISIPTA 2013 ).

  22. Properties of the Operators of Composition Differences ◮ IPFP: for factorizing operator Csisz´ ar’s convergence theorem holds true, it does not hold for Dempster operator ( R. Jirouˇ sek, V. Kratochv´ ıl. On Open Problems Connected with Application of the Iterative Proportional Fitting Procedure to Belief Functions. ISIPTA 2013 ). ◮ Computational complexity: Factorizing operator is much simpler than Demster operator. • factorizing operator ∼ c ⊂ X K ∪ L : c = c ↓ K ⊲ ⊳ c ↓ L ; • Dempster operator ∼ c ⊂ X K ∪ L .

  23. Properties of the Operators of Composition Differences ◮ IPFP: for factorizing operator Csisz´ ar’s convergence theorem holds true, it does not hold for Dempster operator ( R. Jirouˇ sek, V. Kratochv´ ıl. On Open Problems Connected with Application of the Iterative Proportional Fitting Procedure to Belief Functions. ISIPTA 2013 ). ◮ Computational complexity: Factorizing operator is much simpler than Demster operator. • factorizing operator ∼ c ⊂ X K ∪ L : c = c ↓ K ⊲ ⊳ c ↓ L ; • Dempster operator ∼ c ⊂ X K ∪ L . ◮ Computation of conditionals: for µ on X M and X j , X k ∈ M , � ↓ X k , � ν X j = a d ⊲ µ µ ( X k | X j = a ) = where ν X j = a is a one-dimensional bpa on X j having just one focal element { a } ⊂ X j , for which ν X j = a ( { a } ) = 1.

  24. Properties of the Operators of Composition Conditional independence lemma For commonality function θ ( X , Y , Z ) the conditional independence relation X ⊥ ⊥ θ Y | Z holds true iff θ ( X , Y , Z ) = θ ( X , Z ) d ⊲ π ( Y , Z ) .

  25. Properties of the Operators of Composition Conditional independence lemma For commonality function θ ( X , Y , Z ) the conditional independence relation X ⊥ ⊥ θ Y | Z holds true iff θ ( X , Y , Z ) = θ ( X , Z ) d ⊲ π ( Y , Z ) . Factorization Lemma For bpa µ ( X , Y , Z ) there exist functions φ : 2 X × Z − ψ : 2 Y × Z − → R + , → R + , such that if a = a ↓ X × Z ⊲ φ ( a ↓{ X , Z } ) · ψ ( a ↓{ Y , Z } ) ⊳ a ↓ Y × Z � µ ( a ) = 0 otherwise iff µ ( X , Y , Z ) = µ ( X , Z ) f ⊲ µ ( Y , Z ) .

  26. Conclusions - An Open Problem ◮ It seems recommendable to use the factorizing operator of composition for the efficient representation of multidimensional compositional models,

Recommend


More recommend