Abstract
Approximation operations induced by coverings are reinterpreted through a set of four “constructors” defined by simple logical formulas. The very logical definitions of the constructors make it possible to readily understand the properties of such operators and their meanings.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The members of U will be usually denoted by g after the German term Gegenstand which means an object before interpretation, while M is after Merkmal, which means “property”. In Formal Topology, M is thought of as a set of abstract neighborhoods. This interpretation will be used later on in the paper.
- 2.
R(A) and \(R^\smile (B)\) are also called the the left Peirce product of R and A, and, respectively, the right Peirce product of R (left Peirce product of \(R^\smile \)) and B.
- 3.
Often a lower adjoint is called “left adjoint” and an upper adjoint “right adjoint”. We avoid the terms “right” and “left” because they could make confusion with the position of the arguments of the operations of binary relations. For the general notion of adjoint functors see for instance [3]. For Galois connections induced by binary relations a classic reference is [19]. For the present use in Rough Set Theory see [23] or [25].
- 4.
It is worth noticing that there are constructive logics between Intuitionistic and Classical logics such that the opposite of the above implication holds if the premise is a negated formula (see [14]).
- 5.
In these works \({\mathcal N}(\mathbf{P})\) is denoted as \({\mathcal N}(U)\) and instead of \({\mathcal Z}\) the entire powerset \(\wp (U)\) is considered. The present is a slight generalization.
- 6.
- 7.
If R is not serial and \(R(g)=\emptyset \), then \(\emptyset \) does not belong to \({\mathcal N}_g\). Otherwise stated, \(\emptyset \) is different from \(\{\emptyset \}\). 0 does not hold if there exists \(g\in G\) such that \(\langle g,\emptyset \rangle \in R\).
- 8.
In [22] \({\mathcal N}_*(U)\) is denoted as \({\mathcal N}_{F(R)}(U)\), and \({\mathcal N}^*_x\) as \({\mathcal N}^R_x\).
- 9.
A wider reference about covering-based approximation operators and the scientific literature about the topic can be found in Sect. 5 of [33].
- 10.
The original definition of \((uC)_7\) is \((lC)_1(X)\cup (\bigcup \{n(x):x\in X\cap -(lC)_1(X)\})\).
- 11.
Actually, from Facts 3.(iii), Corollary 1 and Lemma 11.(5) and (6) one trivially derives that in any SRS P with R a preorder: \(\mathbf{S}_{\langle i\rangle }(\mathbf{P})=\mathbf{S}_{[e]\langle i\rangle }(\mathbf{P})=\mathbf{S}_{[e]}(\mathbf{P})= \mathbf{S}_{\langle i\rangle [e]}(\mathbf{P})\); \(\mathbf{S}_{\langle e\rangle }(\mathbf{P})=\mathbf{S}_{[i]\langle e\rangle }(\mathbf{P})=\mathbf{S}_{[i]}(\mathbf{P})= \mathbf{S}_{\langle e\rangle [i]}(\mathbf{P})\).
- 12.
In general, from Facts 3, if R is a preorder then the set of fixpoints of the constructors \(\langle \cdot \rangle \) and \([\cdot ]\) coincides with the sets of fixpoint of their derived operators \(\langle \cdot \rangle [\cdot ]\) and \([\cdot ]\langle \cdot \rangle \) (where the directions, intension or intension, alternate). Since the sets of fixpoints of the derived operators form distributive lattices, the same happens for the constructors.
References
Bartol, W., Miró, J., Pióro, K., Rosselló, F.: On the coverings by tolerance classes. Inf. Sci. 166, 193–211 (2004)
Bell, J.L.: Orthologic, forcing, and the manifestation of attributes. In: Chong, C.T., Wicks, M.J. (eds.) Southeast Asian Conference on Logic, pp. 13–36. Elsevier Science Publisher, North-Holland (1983)
Blyth, T.S., Janowitz, M.F.: Residuation Theory. Pergamon Press, Oxford (1972)
Bonikowski, Z., Bryniarski, E., Wybraniec-Skardowska, U.: Extensions and intensions in the rough set theory. Inf. Sci. 107, 149–167 (1998)
Čech, E.: Topological Spaces. Wiley, Hoboken (1966)
Düntsch, I., Gegida, G.: Modal-style operators in qualitative data analysis. In: Proceedings of the IEEE International Conference on Data Mining, pp. 155–162 (2002)
Ghanim, M.H., Mustafa, H.I., Abd El Aziz, S.: On lower and upper intension order relations by different cover concepts. Inf. Sci. 181, 3723–3734 (2011)
Greco, S., Matarazzo, B., Słowinski, R.: Algebra and topology for dominance-based rough set approach. In: Ras, Z.W., Tsay, L.-S. (eds.) Advanced in Intelligent Systems. SCI, vol. 265, pp. 43–78. Springer, Heidelberg (2010)
Huang, A., Zhu, W.: Topological characterizations for three covering approximation operators. In: Ciucci, D., Inuiguchi, M., Yao, Y., Ślęzak, D., Wang, G. (eds.) RSFDGrC 2013. LNCS, vol. 8170, pp. 277–284. Springer, Heidelberg (2013)
Järvinen, J.: Lattice theory for rough sets. In: Peters, J.F., Skowron, A., Düntsch, I., Grzymała-Busse, J.W., Orłowska, E., Polkowski, L. (eds.) Transactions on Rough Sets VI. LNCS, vol. 4374, pp. 400–498. Springer, Heidelberg (2007)
Järvinen, J., Pagliani, P., Radeleczki, S.: Information completeness in nelson algebras of rough sets induced by quasiorders. Stud. Log. 101(5), 1073–1092 (2013)
Järvinen, J., Radeleczki, S.: Representation of Nelson algebras by rough sets determined by quasiorders. Algebra Univers. 66, 163–179 (2011)
Järvinen, J., Radeleczki, S.: Tolerances induced by irredundant coverings. Fundamenta Informaticae 137, 341–353 (2015)
Kreisel, G., Putnam, H.: Eine Unableitbarkeitsbeweismethode für den intuitionistischen Aussegenkalkül. Archiv für mathematische Logik und Grundlagenferschung 3, 74–78 (1957)
Kumar, A., Banerjee, M.: Definable and rough sets in covering-based approximation spaces. In: Li, T., Nguyen, H.S., Wang, G., Grzymala-Busse, J., Janicki, R., Hassanien, A.E., Yu, H. (eds.) RSKT 2012. LNCS, vol. 7414, pp. 488–495. Springer, Heidelberg (2012)
Langeron, C., Bonnevay, S.: Une approach pretopologique pour la structuration. Document de Recherche \(N^o\) 1999 - 7. Centre de reserches economiques de l’Université de Saint Etienne
Li, Q., Zhu, W.: Lattice structures of fixed points of the lower approximations of two types of covering-based rough sets. http://arxiv.org/abs/1209.5569
Nagarajan, E.K.R., Umadevi, D.: A method of representing rough sets system determined by quasi orders. Order 30(1), 313–337 (2013)
Ore, O.: Galois connexions. Trans. Am. Math. Soc. 55, 493–513 (1944)
Pagliani, P.: Concrete neighbourhood systems and formal pretopological spaces (draft). In: Presented at the Calcutta Logical Circle Conference on Logic and Artificial Intelligence, Calcutta, 13–16 October 2003
Pagliani, P.: Pretopology and dynamic spaces. In: Wang, G., Liu, Q., Yao, Y., Skowron, A. (eds.) RSFDGrC 2003. LNCS, vol. 2639, pp. 146–155. Springer, Heidelberg (2003). Extended Version in Fundamenta 59(2–3), 221–239 (2004)
Pagliani, P.: The relational construction of conceptual patterns - tools, implementation and theory. In: Kryszkiewicz, M., Cornelis, C., Ciucci, D., Medina-Moreno, J., Motoda, H., Raś, Z.W. (eds.) RSEISP 2014. LNCS, vol. 8537, pp. 14–27. Springer, Heidelberg (2014)
Pagliani, P., Chakraborty, M.K.: Information Quanta and approximation spaces. I: Non-classical approximation operators. In: Proceedings of the IEEE International Conference on Granular Computing, Beijing, China, 25–27 July 2005, vol. 2, pp. 605–610. IEEE, Los Alamitos, July 2005
Pagliani, P., Chakraborty, M.K.: Formal topology and information systems. In: Peters, J.F., Skowron, A., Düntsch, I., Grzymała-Busse, J.W., Orłowska, E., Polkowski, L. (eds.) Transactions on Rough Sets VI. LNCS, vol. 4374, pp. 253–297. Springer, Heidelberg (2007)
Pagliani, P., Chakraborty, M.K.: A Geometry of Approximation. Trends in Logic, 27th edn. Springer, Heidelberg (2008)
Qin, K., Gao, Y., Pei, Z.: On covering rough sets. In: Yao, J.T., Lingras, P., Wu, W.-Z., Szczuka, M.S., Cercone, N.J., Ślȩzak, D. (eds.) RSKT 2007. LNCS (LNAI), vol. 4481, pp. 34–41. Springer, Heidelberg (2007)
Restrepo, M., Cornelis, C., Gmez, J.: Duality, conjugacy and adjointness of approximation operators in covering-based rough sets. Int. J. Approximate Reasoning 55(1), 469–485 (2014)
Sambin, G.: Intuitionistic formal spaces - a first communication. In: Skordev, D. (ed.) Mathematical Logic and Its Applications, pp. 187–204. Plenum Press, New York (1987)
Sambin, G., Gebellato, S.: A preview of the basic picture: a new perspective on formal topology. In: Altenkirch, T., Reus, B., Naraschewski, W. (eds.) TYPES 1998. LNCS, vol. 1657, pp. 194–208. Springer, Heidelberg (1999). doi:10.1007/3-540-48167-2_14
Thuan, N.D.: Covering rough sets from a topological point of view. Int. J. Comput. Theor. Eng. 1(5), 606–609 (2009)
Vickers, S.: Topology via Logic. Cambridge Tracts in Theoretical Computer Science, vol. 5. Cambridge University Press, Cambridge (1989)
Yao, Y.Y., Chen, Y.H.: Rough set approximations in formal concept analysis. In: Dick, S., Kurgan, L., Pedrycz, W., Reformat, M. (eds.) Proceedins of 2004 Annual Meeting of the North American Fuzzy Information Processing Society, (NAFIPS 2004). IEEE Catalog Number: 04TH8736, pp. 73–78 (2004)
Yao, Y., Yao, B.: Covering based rough set approximations. Inf. Sci. 200, 91–107 (2012)
Zakowski, W.: Approximations in the space \((U,\Pi )\). Demonstratio Mathematica 16, 761–769 (1983)
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
A Appendix
A Appendix
The following result is the basis of the calculus of constructors by means of Boolean matrices:
Duality between the constructors. In any SRS P, \(\langle e\rangle \) and [e] are dual; \(\langle i\rangle \) and [i] are dual.
Proof
\(-R^\smile (-A)=\{x:R(x)\subseteq A\}\) (aka: \(-\langle e\rangle (-A)=[e](A)\)).
Same for the intensional pair of constructors. \(\Box \)
Direct proof of Lemma 2.(2). The proof comes straightforwardly from (14). Indeed, \(\bigcap \{-X\in {\mathcal Z}:X\cap A=\emptyset \}=\bigcap \{-R^\smile (m):R^\smile (m)\subseteq -A\}= \{g:\forall m((R^\smile (m)\subseteq -A)\Longrightarrow g\notin R^\smile (m))\}= \{g:\forall m(g\in R^\smile (m)\Longrightarrow R^\smile (m)\nsubseteq -A)\}= \{g:\forall m(m\in R(g)\Longrightarrow R^\smile (m)\cap A\ne \emptyset )\}\). From Lemma 1, the latter set equals \(\{g:\forall m(m\in R(g)\Longrightarrow m\in R(A))\}=\{g:R(g)\subseteq R(A)\}\).
Direct proof of Theorem 12.(7). Since R is a preorder, for any \(X\subseteq U\), \([e](X)\subseteq X\) so that \([e]\langle e\rangle [e](X)\subseteq [e]\langle e\rangle (X)\). Suppose now \(x\in [e]\langle e\rangle (X)\). Then \(x\in [e](R^\smile (X))\) so that \(R(x)\subseteq R^\smile (X)\). If \(x\notin [e]\langle e\rangle [e](X)\) then \(R(x)\nsubseteq R^\smile (\{y:R(y)\subseteq X\})\). Therefore, \(\exists z\) such that \(z\in R(x)\) and \(z\notin R^\smile (\{y:R(y)\subseteq X\})\). Therefore, \(\forall y(R(y)\subseteq X\Longrightarrow y\notin R(z))\), so that \(\forall y(y\in R(z)\Longrightarrow R(y)\nsubseteq X)\). But R is transitive. So \(R(y)\subseteq R(z)\). It follows \(R(z)\nsubseteq X\), which leads to a contradiction because \(z\in R(x)\) so that \(R(z)\subseteq R(x)\) which implies \(R(x)\nsubseteq X\).
Direct proof of Lemma 7.(2):
Direct proof of Lemma 7.(5):
Moreover, \((uC)_6(A)=\bigcup \{K:K\cap A\ne \emptyset \}\). Thus \(x\in (uC)_6(A)\) iff \(\exists K(x\in K\wedge K\cap A\ne \emptyset )\) iff \(\exists K(x\in R^\smile (K)\wedge R^\smile (\{K\})\cap A\ne \emptyset )\) iff \(\exists K(x\in \langle e\rangle (K)\wedge K\in \langle i\rangle (A))\) iff \(x\in \langle e\rangle \langle i\rangle (A)\).
Direct proof of Lemma 12: \(\langle i\rangle (X)=X\) iff \(R(X)=X\) iff \(\bigcup \{R(x):x\in X\}=X\) which implies \(\forall x\in X(R(x)\subseteq X)\). Let \(y\notin X\). Since R is reflexive, \(x\in R(x)\), so that \(R(x)\nsubseteq X\). In sum, \(x\in X\Longrightarrow R(x)\subseteq X\) and \(x\notin X\Longrightarrow R(x) \nsubseteq X\). We conclude that \(x\in X\) iff \(R(x)\subseteq X\) which amounts to saying \(R(X)=\{x:R(x)\subseteq X\}=[e](X)\). Notice that seriality, trivially, is not enough. Indeed seriality does not prevent from the existence of a \(g\in X\) such that \(R(g)\cap X=\emptyset \) - think of the relation \(\{\langle a,b\rangle ,\langle b,b\rangle \}\) on the set \(\{a,b\}\) and put \(X=\{a\}\).
Alternative proof of Theorem 13 (point (3) of Corollary 11 below):
Lemma 18
Let C be unary and \(\mathbf{C}^\blacktriangledown =\{K\in \mathbf{C}:\exists x\wedge md(x)=\{K\}\}\). Then \(R_{\mathbf{C}^\blacktriangledown }=R_\mathbf{C}\).
Proof
Since C is unary, \(\mathbf{P}(\mathbf{C^\blacktriangledown })=\langle U, C^\blacktriangledown , R^\blacktriangledown \rangle \), where \(R^\blacktriangledown \) is R restricted to \(U\times \mathbf{C}^\blacktriangledown \), is a specialised covering of U. So, let us assume \(y\in R_{\mathbf{C}^\blacktriangledown }(x)\). Therefore, \(\forall K\in \mathbf{C}^\blacktriangledown (x\in K\Longrightarrow y\in K)\). In particular this occurs for \(\{K\}=md(x)\). Then for all \(K\in \mathbf{C}(x\in K\Longrightarrow y\in K)\) so that \(y\in R_\mathbf{C}(x)\). Thus \(R_{\mathbf{C}^\blacktriangledown }\subseteq R_\mathbf{C}\). Vice-versa, since \(\mathbf{C}^\blacktriangledown \subseteq \mathbf{C}\), \(R_\mathbf{C}\subseteq R_{\mathbf{C}^\blacktriangledown }\). \(\Box \)
Corollary 11
For any unary covering C:
-
(1)
\(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))=\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}^\blacktriangledown ))\);
-
(2)
\(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))=\mathbf{S}_{\mathcal C}(\mathbf{P}(R_{\mathbf{C}^\blacktriangledown }))\);
-
(3)
\(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))\) is a topological space.
Proof
(1) From Theorem 9: \(\preceq _{{int}^{\mathbf{P}(\mathbf{C})}}=R_{\mathbf{C}}\) and \(\preceq _{{int}^{\mathbf{P}(\mathbf{C^\blacktriangledown })}}=R_{\mathbf{C}^\blacktriangledown }\). But from Lemma 18, \(R_\mathbf{C}=R_{\mathbf{C}^\blacktriangledown }\). Therefore, \(\preceq _{{int}^{\mathbf{P}(\mathbf{C})}}=\preceq _{{int}^{\mathbf{P}(\mathbf{C^\blacktriangledown })}}\) so that \(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))=\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}^\blacktriangledown ))\). (2) From Corollary 7, \(int^{\mathbf{P}(\mathbf{C}^\blacktriangledown )}=int^{\mathbf{P}(R_{\mathbf{C}^\blacktriangledown })}\), because \(\mathbf{P}(\mathbf{C}^\blacktriangledown )\) is a specialised covering whenever C is unary. In view of (1) one concludes \(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))=\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}^\blacktriangledown ))= \mathbf{S}_{\mathcal C}(\mathbf{P}(R_{\mathbf{C}^\blacktriangledown }))\). (3) is obtained from (1) and Theorem 13. \(\Box \)
Corollary 11.(3) is a proof, by means of the constructor approach, of Theorem 12 of [17].
The following Lemma helps clarifying the isomorphism stated above:
Lemma 19
\(\forall K\in \mathbf{C}\cap - \mathbf{C}^\blacktriangledown \), \(\exists {\mathcal K}\subseteq \mathbf{C}^\blacktriangledown \) such that \(K=\bigcup {\mathcal K}\).
Proof
Let \(K\in \mathbf{C}\cap - \mathbf{C}^\blacktriangledown \) and \(x\in K\). Clearly, \(K_x\varsubsetneq K\), for minimality of \(K_x\). Therefore, \(\bigcup \{K_x:x\in K\}\subseteq K\). But by definition, \(x\in K_x\). Therefore, \(\bigcup \{K_x:x\in K\}=K\). \(\Box \)
Example 11
Let \(\mathbf{C}=\{\{a\},\{b\},\{a,b\},\{a,c\}\}\). \(\mathbf{C}^\blacktriangledown =\{\{a\},\{b\},\{a,c\}\}\).
Theorem 16
If C is a representative covering of a set U, then C is the set of coprime elements of \(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))\).
Proof
Since C is representative, for all \(K\in \mathbf{C}\), \(K=K_x\) for some \(x\in U\). Therefore, for all \(K'\varsubsetneq K\), \(x\notin K'\), by definition of \(K_x\). It follows that if \({\mathcal K}\subseteq \mathbf{C}\) and \(K\notin {\mathcal K}\), then \(K\ne \bigcup {\mathcal K}\). \(\Box \)
Justification of the definition of an “irredundant covering”. We recall that in [13] a covering C of a set U is called “irredundant” if for any \(K\in \mathbf{C}\), \(\mathbf{C}\cap -\{K\}\) is no longer a covering of U.
Theorem 17
C is an irredundant covering if and only if it is representative and reduced.
Proof
(A) If C is redundant, then either it is not reduced or it is not representative. Assume \(K\in \mathbf{C}\) is redundant. Then for all \(x\in K\), there is a \(K'\in \mathbf{C}\) such that \(x\in K'\). If \(K\subseteq K'\) or \(K'\subseteq K\), then C is not reduced. If it is reduced, then \(K\cap K'\notin \mathbf{C}\), because \(K\cap K'\subseteq K\) and \(K\cap K'\subseteq K'\). It follows that for all \(x\in K\), \(K\ne K_x\). Hence K is not represented. (B) If C is either non-reduced or non-representative, then it is redundant. Trivially, if C is not reduced, and \(K\varsubsetneq K'\), then K is redundant. Suppose C is reduced but there exists \(K\in \mathbf{C}\) which is not represented. Then for all \(x\in K\), \(K\ne K_x\). It follows that for all \(x\in K\), there is a \(K'\ne K\) such that \(x\in K'\). We conclude that K is redundant. \(\Box \)
The composition of two relations \(R\subseteq X\times Y\) and \(Z\subseteq Y\times W\), is defined as \(R\otimes Z=\{\langle x,w\rangle : \exists y\in Y(\langle x,y\rangle \in R\wedge \langle y,w\rangle \in Z\}\).
In [13] it is stated that if R is a preorder, then \(R^\smile \otimes R\), is a tolerance relation. We know that \(R_\mathbf{C}\) is a preorder. Now we prove that if in addition C is reduced, then \(R^\smile _\mathbf{C}\otimes R_\mathbf{C}(x)=R_\mathbf{C}(x)\) for all \(x\in U^\blacktriangleleft \). Since \(R^\smile _\mathbf{C}\otimes R_\mathbf{C}(x)\) is a tolerance relation, as a corollary we have that if C in addition is specialised then \(R_\mathbf{C}\) is an equivalence relation. From that we can deduce that \(\mathbf{S}_{int}(\mathbf{P}(\mathbf{C}))\) is a Boolean algebra (i.e. Theorem 15).
Lemma 20
For any covering C of U, for all \(x,y\in U\), \(\langle x,y\rangle \in R^\smile _\mathbf{C}\otimes R_\mathbf{C}\) iff \(\exists z(\langle z,x\rangle \in R_\mathbf{C}\wedge \langle z,y\rangle \in R_\mathbf{C})\).
The proof is trivial.
Lemma 21
For any covering C of U, \(R^\smile _\mathbf{C}\otimes R_\mathbf{C}\supseteq R_\mathbf{C}\).
Proof
Let \(\langle x,y\rangle \in R_\mathbf{C}\). From reflexivity, \(\langle x,x\rangle \in R_\mathbf{C}\), so that from Lemma 20, \(\langle x,y\rangle \in R^\smile _\mathbf{C}\otimes R_\mathbf{C}\). \(\Box \)
Theorem 18
For any reduced covering C of U, \(R^\smile _\mathbf{C}\otimes R_\mathbf{C}(x)=R_\mathbf{C}(x)\) for all \(x\in U^\blacktriangleleft \).
Proof
Suppose \(x\in U^\blacktriangleleft \) and \(\langle x,y\rangle \in R^\smile _\mathbf{C}\otimes R_\mathbf{C}\). Then \(\exists z\) such that \(x,y\in R_\mathbf{C}(z)=n(z)=\bigcap md(z)\). Since x is representative there is a \(K_x\in \mathbf{C}\) and since \(x\in n(x)\), \(K_x\subseteq K\) for all \(K\in md(z)\). But C is reduced, so that it must be \(K=K_x\) for all \(K\in md(z)\). That is, \(md(z)=\{K_x\}\), so that \(R_\mathbf{C}(z)=R_\mathbf{C}(x)\). It follows that \(y\in R_\mathbf{C}(x)\) and one concludes \(R^\smile _\mathbf{C}\otimes R_\mathbf{C}\subseteq R_\mathbf{C}\). In view of Lemma 21 the proof is complete. \(\Box \)
Rights and permissions
Copyright information
© 2016 Springer-Verlag GmbH Germany
About this chapter
Cite this chapter
Pagliani, P. (2016). Covering Rough Sets and Formal Topology – A Uniform Approach Through Intensional and Extensional Constructors. In: Peters, J., Skowron, A. (eds) Transactions on Rough Sets XX. Lecture Notes in Computer Science(), vol 10020. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-662-53611-7_4
Download citation
DOI: https://doi.org/10.1007/978-3-662-53611-7_4
Published:
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-662-53610-0
Online ISBN: 978-3-662-53611-7
eBook Packages: Computer ScienceComputer Science (R0)