0% found this document useful (0 votes)
32 views

Measuring Incoherence

The document discusses measuring incoherence in extended multi-adjoint logic programs. It introduces multi-adjoint lattices and extended logic programs that allow for multiple types of strong negation. It defines coherence for interpretations and programs, and proposes measuring the incoherence of pairs of opposite literals by determining the minimum amount of information that must be removed from their interpretations to make the pair coherent. The paper aims to generalize previous work on measuring incoherence in residuated logic programs to the current multi-adjoint framework.

Uploaded by

Ali Cauchy
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
32 views

Measuring Incoherence

The document discusses measuring incoherence in extended multi-adjoint logic programs. It introduces multi-adjoint lattices and extended logic programs that allow for multiple types of strong negation. It defines coherence for interpretations and programs, and proposes measuring the incoherence of pairs of opposite literals by determining the minimum amount of information that must be removed from their interpretations to make the pair coherent. The paper aims to generalize previous work on measuring incoherence in residuated logic programs to the current multi-adjoint framework.

Uploaded by

Ali Cauchy
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 27

Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended

programs Conclusions

On the Measure of Incoherent Information in Extended Multi-Adjoint Logic Programs


Nicol as Madrid Manuel Ojeda-Aciego
Univ de M alaga (Spain)

Ostrava Univ (Czech Republic)

April 12, 2013

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Aims of this research


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

We are studying the introduction of two kind of negations into fuzzy frameworks: Default negation: This negation enables non-monotonic reasoning and is introduced by generalizing the semantics of Answer sets and Equilibrium logic. Strong negation: Including this kind of negation into positive fuzzy theories does not aect to the monotonicity of them. In this sense, it is necessary to provide a suitable generalization of the notion of consistency.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Overview
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

In this work we consider several strong negations. So: We start by giving the notion of extended multi-adjoint logic programs. To deal with strong negations we recall (and extend) the notion of coherence given on residuated logic programs. Subsequently, we measure the incoherence by:
focusing rstly on measuring the incoherence on interpretations and later by extending the measures dened on interpretations to extended multi-adjoint logic programs.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Preliminaries
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence 2 Incoherence on Interpretations Incoherence on extended programs Conclusions 3

Denition A multi-adjoint lattice L is a tuple (L, , i , i ) such that:


1

(L, ) is a complete bounded lattice, with top and bottom elements 1 and 0. for all i , for all x the equations 1 i x = x i 1 = x hold. for all i , and for all x , y , z L the tuple (i , i ) forms an adjoint pair, i.e. z (x i y ) i y i z x .

A negation operator on L is any decreasing mapping n : L L satisfying n(0) = 1 and n(1) = 0.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Syntax
Literals
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

The language considered in this approach is propositional. Thus, given L a multi-adjoint lattice with a nite set of negation operators {j }j J : We denote by the set of propositional symbols We dene the set of literals by Lit = {j p | p and j J }

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Syntax
Extended multi-adjoint logic program
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Given a multi-adjoint lattice with negations (L, , i , i , j ), an extended multi-adjoint logic program P is a nite set of weighted rules of the form F ; satisfying the following conditions: F is a formula of the form i B where is a literal (called the head of F ) and B (called the body of F ) is built from literals 1 , . . . , n and operators i . the weight is an element of the underlying multi-adjoint lattice L.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Semantics
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Let L = (L, ) be a bounded lattice, an L-interpretation is a mapping I : Lit L. The domain of the interpretation is the set of literals, and it can be lifted to any rule by homomorphic extension: Denition (Model) We say that I satises a rule i B ; if and only if I (B ) i I ( ) or, equivalently, I ( i B ). Finally, I is a model of P if it satises all rules in P.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

The immediate consequences operator


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

The immediate consequences operator dened on positive multi-adjoint logic programs can be applied straightforwardly to extended programs. We recall its denition below: Denition Let P be an extended multi-adjoint logic program and let I be an L-interpretation. The immediate consequence operator of I wrt P is the L-interpretation dened by TP (I )( ) = sup{I (B ) i : i B ; P}

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

The least model


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

The immediate consequences operator TP is monotonic. By the Knaster-Tarski x-point theorem, TP has a least x-point; lfp (TP ). lfp (TP ) coincides with the least model of P. The least model semantics of an extended multi-adjoint logic program P is given by the lfp (TP ). However, one has to take into account the possible interaction between opposite literals. For this purpose we dene the notion of coherence.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Coherence
The denition
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Let L be a multi-adjoint lattice. An L-interpretation I is coherent if the inequality I (i ) i I ( ) holds for every literal and all strong negation i L. The notion of coherence coincides with consistency in the classical framework. It only depends on negation operators. It allows to handle missing information (i.e I such that I ( ) = 0 for all Lit is always coherent).

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Coherence
Extension to multi-adjoint programs
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Let P be an extended multi-adjoint logic program, we say that P is coherent if its least model is coherent. Theorem An extended multi-adjoint logic program is coherent if and only if it has at least one coherent model.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Incoherent pair of literals


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Let L be a multi-adjoint lattice and let I be an interpretation. We say that ( , i ) is coherent w.r.t. I if and only if the inequality I (i ) i I ( ) holds. Otherwise the pair ( , i ) is called incoherent. However, it is convenient to provide degrees of incoherence. Consider the following two interpretations on {p , p }: I1 (p ) = 0.5 I1 ( p ) = 0.6 I2 (p ) = 1 I2 ( p ) = 0.9

and the usual negation (x ) = 1 x to determine the coherence. Certainly, the pair (p , p ) seems to be more incoherent w.r.t. I2 than w.r.t. I1 .
Manuel Ojeda-Aciego Incoherence in Multi-Adjoint Programs

Information measure
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

We propose to assign a value to each element in the lattice corresponding to the inherent information it contains. Denition Let (L, ) be a lattice, an information measure is an operator m : L R+ such that the following holds for all x , y L: m is monotonic. m(x ) = 0 if and only if m(x ) = 0. m(sup(x , y )) m(x ) + m(y ) m(inf(x , y )). Note that the third item imposes no restriction if the lattice is linearly ordered.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

The set of coherent pairs w.r.t. i


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Hereafter we will assume that our multi-adjoint lattices have an associated information measure m. In order to measure the degree of incoherence of a pair ( , i ) w.r.t. I we focus on the minimal amount of information we have to remove from I ( ) and I (i ) in order to recover the coherence of the pair ( , i ). Denition Let i be a negation operator. The set of coherent pairs w.r.t. i is the set: i = {(x , y ) L L : y i (x ) }

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Measure of incoherence for pairs of opposite literals


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition (I L ) We dene the measure of incoherence of a pair ( , i ) w.r.t. one interpretation I (denoted by I L (( , i ); I )) as follows:
(x ,y )i (x ,y )(I ( ),I (i ))

inf

m(I ( )) m(x ) + m(I (i )) m(y )

where the ordering within i is considered componentwise. I L (( , i ); I ) measures how much information we have to remove from I ( ) and I (i ) in order to obtain a coherent pair of literals ( , i ) w.r.t. I .

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

A specic case
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

The denition of I L (( , i ); I ) collapses to an extremely simple and intuitive form in the specic case of the unit interval L = [0, 1], the information measure induced by the Euclidean norm, and i (x ) = 1 x : 0 if ( , i ) is coherent I L (( , ); I ) = i I (i ) i I ( ) otherwise

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Measure of incoherence for literals


Denition of coherent literals
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Here, we dene a general degree of incoherence by considering all possible negations of a literal. Denition Let L be a multi-adjoint lattice and let I be an interpretation. We say that a literal is coherent w.r.t. I if and only if the inequality I (i ) i I ( ) holds for all strong negation i L. Otherwise the literal is called incoherent. Denition Given a multi-adjoint lattice with negations {1 , 2 , . . . , n }, we dene the set of coherent tuples w.r.t. L as L = {(x , y1 , . . . , yn ) Ln+1 : yi i (x ) }
Manuel Ojeda-Aciego Incoherence in Multi-Adjoint Programs

Measure of incoherence I G ( ; I )
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Now, we dene a measure of incoherence for literals, similar to I L , but by considering the set L instead of i . Formally, Denition We dene the measure of incoherence I G ( ; I ) by:
n (x ,yi )L x I ( ) yi I (i )

inf

m(I ( )) m(x ) +
i

m(I (i )) m(yi )

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Relationship between the measures I L and I G


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Proposition Let L be a multi-adjoint lattice and let I be an interpretation. Then for all Lit and i L we have that: I L (( , j ); I ) I G ( ; I ) I L (( , i ); I )
j L

We need to remove less information from I if we deal directly with the incoherence of all negated literals {i }i of than if we deal independently with each incoherent pair ( , i ). Corollary Let L be a multi-adjoint lattice, let I be an interpretation and let be a literal. Then, I G ( ; I ) = 0 if and only if I L (( , i ); I ) = 0 for all pair ( , i )
Manuel Ojeda-Aciego Incoherence in Multi-Adjoint Programs

Properties of I L and I G
Null measure
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Proposition If the literal (resp. the pair ( , i )) is coherent w.r.t. I then I G ( ; I ) = 0 (resp. I L (( , i ); I ) = 0). We can ensure the equivalence between coherence and null measure of incoherence in the following frameworks: Whether the multi-adjoint lattice is nite and the information measure used is injective. Whether the multi-adjoint lattice is the unit interval [0, 1] and the information measure is continuous and injective.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Properties of I L and I G
Monotonicity and bounds
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Proposition Let I J be two L-interpretations. Then I L (( , i ); I ) I L (( , i ); J ) for all pair of literals ( , i ). I G ( ; I ) I G ( ; J ) for all literal . The following proposition shows that I L (( , i ); I ) is bounded by the inherent information in I ( ) and I (i ): Proposition Let I be an L-interpretation, then I L (( , i ); I ) min m(I (i )), m(I ( ))
Manuel Ojeda-Aciego Incoherence in Multi-Adjoint Programs

The average number of incoherences


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Let P be an extended program. We denote the number of incoherent literals w.r.t. MP (least model of P) as N I (P) and the number of incoherent pairs of opposite literals w.r.t. MP as N IP (P). So we can consider the measures of incoherence L (P) and I G (P) as: I1 1
L I1 (P) =

N IP (P) |LitP | |P | N I (P) |LitP |

(1)

G I1 (P) =

(2)

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

The maximal size of incoherence


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

L and I G focus on estimating the maximal size The measures I2 2 of incoherence in P.

Denition Given an extended program P, we consider:


L I2 (P) = max {I L (( , i ); MP )} i LitP

(3)

and
G I2 (P) = max {I G ( ; MP )} LitP

(4)

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

The average size of incoherence


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

L and I G can be dened by focusing on The measures I3 3 estimating the average size of incoherence in P.

Denition Given an extended program P, we can consider:


L I3 (P) = i LitP i

I L (( , i ); MP ) L LitP I (( , i ); I )
LitP

(5)

and
G I3 (P) =

I G ( ; MP ) G LitP I ( ; I )

(6)

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Weighted measures of incoherence


Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

Denition Given an extended program P and a set of weights (with the form {( ,i ) }, resp. { }, with LitP ) we consider:
L I4 (P; {( ,i ) })

=
i LitP

, i )

I L (( , i ); MP )

(7)

and
G I4 (P; { }) = LitP

I G ( ; MP )

(8)

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Conclusions
Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

We have extended the semantics of multi-adjoint logic programs to allow the use of several strong negations. We have recalled the notion of coherent L-interpretations. We have measure the incoherence on interpretations by:
determining the incoherence in each pair ( , i ) measuring the incoherence related to each literal

Finally we have extended the measures of incoherence dened on interpretations to extended programs by:
determining the average number of incoherences. measuring the greatest degree of incoherence. quantifying the average degree of incoherences. assigning weights to literals.

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

Incoherence in Multi-Adjoint Programs Manuel Ojeda-Aciego Introduction Preliminaries Least model and coherence Incoherence on Interpretations Incoherence on extended programs Conclusions

On the Measure of Incoherent Information in Extended Multi-Adjoint Logic Programs


Nicol as Madrid Manuel Ojeda-Aciego
Univ de M alaga (Spain)

Ostrava Univ (Czech Republic)

April 12, 2013

Manuel Ojeda-Aciego

Incoherence in Multi-Adjoint Programs

You might also like