Elsevier

Fuzzy Sets and Systems

Volume 336, 1 April 2018, Pages 116-147
Fuzzy Sets and Systems

Single axioms for lower fuzzy rough approximation operators determined by fuzzy implications

https://doi.org/10.1016/j.fss.2017.05.007Get rights and content

Abstract

This paper further investigates axiomatic characterizations of lower fuzzy rough approximation operators determined by a semicontinuous fuzzy implication I. The I-lower fuzzy rough approximation operators on a general fuzzy relation can be characterized by only one axiom. The axiomatic characterizations of I-lower fuzzy rough approximation operators are studied in two approaches, which characterize abstract I-lower fuzzy rough approximation operators with ordinary fuzzy operations on fuzzy sets and product operation, respectively. Considering the properties of fuzzy implications, different axioms are proposed for I-lower fuzzy rough approximation operators, which are determined by an S-implication, an R-implication and a fuzzy implication satisfying left neutrality property and regular property, respectively. When the fuzzy relation is special, such as serial, reflexive, symmetric and so on, I-lower fuzzy rough approximation operators can also be characterized by a single axiom.

Introduction

Rough set theory was proposed by Pawlak [17], [18] to deal with uncertainty due to incompleteness and indiscernibility in information systems. However, Pawlak introduced rough sets with equivalence relations, which are too strict to limit the applicabilities of rough sets. The equivalence relation in rough sets was replaced with a general binary relation or a neighborhood operator [3], [33] to expand rough sets. Dubois and Prade [6] combined fuzzy sets and rough sets to further expand rough sets in the fuzzy environment. Moreover, the combinations of fuzzy sets and rough sets were investigated with different fuzzy logic operations and binary fuzzy relations in [2], [15], [16], [19], [20], [26], [28], [34], where fuzzy implications play an important role in the extensions of fuzzy rough sets.

In recent years, there are mainly two different approaches for the study of rough set theory, i.e., the constructive way and axiomatic way. In the constructive approach, lower and upper approximation operators are constructed and investigated on a binary relation. On the contrary, abstract lower and upper approximation operators are characterized by certain axioms in the axiomatic approach. The axiomatic approach in rough set theory has been discussed in many published works [11], [22], [23], [32], [35], [37]. Especially, the minimal independent axiom sets characterizing rough sets have been studied in [12], [13], [31], [36]. Considering rough approximation operators in the fuzzy environment, Liu [12] sought for one axiom for fuzzy rough approximation operators. In [30], Wu et al. characterized (S,T)-fuzzy rough approximation operators [15] by only one axiom, where S and T denote a t-conorm and a t-norm on the unit interval [0,1], respectively. However, t-(co)norms are assumed to be continuous and, t-norms and t-conorms are assumed to be dual with respect to the standard negation in [30], which seem to be too strict to limit the applicabilities of those axiomatic characterizations.

The axiomatic characterizations of lower fuzzy rough approximation operators determined by fuzzy implications have been studied in [16], [27]. For example, Morsi and Yakout [16] investigated the axiomatic characterizations of fuzzy rough sets based on left-continuous t-norms and R-implications. As a left-continuous t-norm and its R-implication form a special residuated lattice, i.e., an MTL-algebra [7], L-fuzzy rough sets were studied in [14], [21], [25] in the axiomatic approach, where L denotes a residuated lattice. Wu et al. [27] investigated the axiomatic characterizations of (I,T)-fuzzy rough sets [19], where I denotes a fuzzy implication. However, the axiomatic systems mentioned above consist of at least two axioms. Wu et al. [29] used one axiom to characterize fuzzy rough approximation operators determined by fuzzy implications, where fuzzy implication I is always assumed to be continuous and satisfy both exchange principle and ordering property. Meanwhile, fuzzy rough approximation operators determined by fuzzy implications cannot be characterized by only one axiom in [29], when fuzzy relation is symmetric. In fact, Baczyński and Jayaram [1] pointed out that a fuzzy implication satisfying continuity and exchange principle is an S-implication with a continuous fuzzy negation. Moreover, an S-implication satisfying ordering property must be an S-implication with a strong fuzzy negation. Hence the axiomatic characterizations of I-lower fuzzy rough approximation operators by a single axiom in [12], [29], [30] are limited into a special case, where fuzzy implication is a continuous S-implication. In [10], lower fuzzy rough approximation operator of an L-fuzzy rough set was characterized by a single axiom, when fuzzy implication in L satisfies regular property. However, that assumption is too strict for a general R-implication. For example, the R-implication based on minimum does not satisfy regular property. These suggest that single axioms for I-lower fuzzy rough approximation operators should be studied more thoroughly in this context, which further studies the axiomatic characterizations of (I,T)-fuzzy rough sets [19], I-fuzzy rough sets [28] and L-fuzzy rough sets. Moreover, the single axioms for I-lower fuzzy rough approximation operators have axiomatic characterizations of both fuzzy rough sets in [12] and (S,T)-fuzzy rough sets in [30] as special cases.

In this paper, we discuss the axiomatic characterizations of I-lower fuzzy rough approximation operators by a single axiom. Firstly, we investigate single axioms for I-lower fuzzy rough approximation operators determined by specific fuzzy implications, such as S-implications and R-implications. Secondly, we further study single axioms for I-lower fuzzy rough approximation operators determined by fuzzy implications satisfying left neutrality property and regular property instead of specific fuzzy implications. Since both fuzzy rough sets and (S,T)-fuzzy rough sets are special cases of (I,T)-fuzzy rough sets, we derive the conclusions in [12], [29], [30] as special cases of our research.

The content of the paper is organized as follows. In Section 2, we recall some fundamental concepts and related properties of (I,T)-fuzzy rough sets. In Section 3, we characterize I-lower fuzzy rough approximation operators by only one axiom, where fuzzy implication I is an S-implication generated from a right-continuous t-conorm. Section 4 studies single axioms for I-lower fuzzy rough approximation operators determined by R-implications. In Section 5, we discuss the axiomatic characterizations of I-lower fuzzy rough approximation operators determined by fuzzy implications satisfying left neutrality property and regular property. Moreover, fuzzy implication is assumed to satisfy ordering property, when we discuss single axioms for I-lower fuzzy rough approximation operators on a reflexive, T-transitive or T-Euclidean fuzzy relation. In the final section, we present some conclusions of our research.

Section snippets

Preliminaries

Let [0,1] be the unit interval and X be a universe. Then a mapping A:X[0,1] is called a fuzzy set on X. The family of all fuzzy sets on X is denoted as F(X). Let α[0,1]. Then a fuzzy set AF(X) is a constant, while A(x)=α for all xX, denoted as αX. For convenience, we also denote the empty set ∅ and the universe X as 0X and 1X, respectively. Especially, if for all xX,A(x)={1,x=y;0,xy; then A is denoted as 1y. A decreasing function N:[0,1][0,1] is called a fuzzy negation, if it satisfies N(

Axioms for I-lower fuzzy rough approximation operators determined by S-implications

From the view of S-implications, Wu et al. [30] characterized I-lower fuzzy rough approximation operators by only one axiom, where I is an S-implication based on a continuous t-conorm and the standard negation. Moreover, Wu et al. [29] characterized I-lower fuzzy rough approximation operators by a single axiom, when fuzzy implication is continuous and satisfies both exchange principle and ordering property. In [29], I-lower fuzzy rough approximation operators cannot be characterized by one

Axioms for I-lower fuzzy rough approximation operators determined by R-implications based on left-continuous t-norms

Although an R-implication based on a left-continuous t-norm satisfies (EP), (NP) and (OP) [1], Example 2.2 shows that an R-implication based on minimum does not satisfy continuity or regular property. Hence neither the conclusions in [29], [30] nor Section 3 can characterize I-lower fuzzy rough approximation operator determined by a general R-implication. In this section, we study the axiomatic characterizations of I-lower fuzzy rough approximation operators determined by general R-implications

Axioms for I-lower fuzzy rough approximation operators determined by fuzzy implications satisfying left neutrality property and regular property

By Note 2.3, a QL-implication satisfies regular property, when it is generated by a strong fuzzy negation. It is obvious that a QL-implication satisfies (NP). Moreover, Dimuro et al. [4] pointed out that if 0 is the neutral element of grouping function G and N is a strong fuzzy negation, then a G-implication satisfies both (NP) and regular property. From a more general form, we study the single axioms for I-lower fuzzy rough approximation operators determined by fuzzy implications satisfying

Conclusion

Since the axiomatic characterizations of I-lower fuzzy rough approximation operators by only one axiom in [12], [29], [30] focus on continuous S-implications, we further investigate single axioms for I-lower fuzzy rough approximation operators determined by semicontinuous S-implications with continuous fuzzy negations and strict fuzzy negations, respectively. We propose two types of axiomatic characterizations of I-lower fuzzy rough approximation operators. We also study single axioms for I

Acknowledgements

The author would like to thank the Editors and the anonymous reviewers for their valuable comments and suggestions in improving this paper. This research was supported by the National Natural Science Foundation of China (Grant No. 11501334) and the Natural Science Foundation of Shandong Province (Grant No. ZR2015FQ014).

References (37)

Cited by (15)

  • Axiomatic characterizations of L-valued rough sets using a single axiom

    2021, Information Sciences
    Citation Excerpt :

    Considering L being a complete Heyting algebra or a frame with an order-reversing involution, Pang et al. [19,21] systematically investigated axiomatic characterizations of L-rough sets with respect to serial, reflexive, symmetric, transitive, adjoint, Euclidean and mediate L-fuzzy relation as well as their compositions by a single axiom, respectively. For more information on single axiomatic characterization of fuzzy rough sets, we can refer to [11,20,32,37]. The structure of this paper is as follows.

  • New results on single axioms for L-fuzzy rough approximation operators

    2020, Fuzzy Sets and Systems
    Citation Excerpt :

    Although the axiomatic characterizations of fuzzy rough sets based on t-norms and fuzzy implications were investigated in [30], each axiom set in [30] consists of at least two axioms. Considering different fuzzy implications, single axioms for lower fuzzy rough approximation operators determined by fuzzy implications were studied in [28,31]. As Goguen [6] pointed out that the degree of membership may not be always represented by elements in the unit interval, a residuated lattice L becomes the main structure of truth values in many-valued logic [7].

  • L-fuzzifying approximation operators in fuzzy rough sets

    2019, Information Sciences
    Citation Excerpt :

    Using single axioms to characterize fuzzy rough approximation operators has become a hot topic in fuzzy rough set theory. Following this topic, Wu et al. [41–43], Bao et al. [1] and Wang [37] made some discussions. However, fuzzy rough approximation operators in these existing results are all based on fuzzy approximated subsets and fuzzy relations, which is the first kind of fuzzy rough approximation operators.

View all citing articles on Scopus
View full text