1: %% This document created by Scientific Word (R) Version 3.5
2:
3: \documentclass[11pt,thmsb,fleqn]{article}%
4: \usepackage{graphicx}
5: \usepackage{amsmath}%
6: \usepackage{amsfonts}%
7: \usepackage{amssymb}
8: %TCIDATA{OutputFilter=latex2.dll}
9: %TCIDATA{CSTFile=article.cst}
10: %TCIDATA{LastRevised=Tuesday, June 18, 2002 00:46:29}
11: %TCIDATA{<META NAME="GraphicsSave" CONTENT="32">}
12: \newtheorem{theorem}{Theorem}[section]
13: \newtheorem{acknowledgement}[theorem]{Acknowledgement}
14: \newtheorem{algorithm}[theorem]{Algorithm}
15: \newtheorem{axiom}[theorem]{Axiom}
16: \newtheorem{case}[theorem]{Case}
17: \newtheorem{claim}[theorem]{Claim}
18: \newtheorem{conclusion}[theorem]{Conclusion}
19: \newtheorem{condition}[theorem]{Condition}
20: \newtheorem{conjecture}[theorem]{Conjecture}
21: \newtheorem{corollary}[theorem]{Corollary}
22: \newtheorem{criterion}[theorem]{Criterion}
23: \newtheorem{definition}[theorem]{Definition}
24: \newtheorem{example}[theorem]{Example}
25: \newtheorem{exercise}[theorem]{Exercise}
26: \newtheorem{lemma}[theorem]{Lemma}
27: \newtheorem{notation}[theorem]{Notation}
28: \newtheorem{problem}[theorem]{Problem}
29: \newtheorem{proposition}[theorem]{Proposition}
30: \newtheorem{remark}[theorem]{Remark}
31: \newtheorem{solution}[theorem]{Solution}
32: \newtheorem{summary}[theorem]{Summary}
33: \newenvironment{proof}[1][Proof]{\textbf{#1.} }{\ \rule{0.5em}{0.5em}}
34: \setlength{\textwidth}{6.75in}
35: \setlength{\textheight}{9.00in}
36: \setlength{\oddsidemargin}{-.125 in}
37: \setlength{\topmargin}{-.60 in}
38:
39:
40: \begin{document}
41:
42: \title{The Lattice of Fuzzy Intervals \\and Sufficient Conditions for Its Distributivity }
43: \author{Ath. Kehagias}
44: \maketitle
45:
46: \begin{abstract}
47: Given a \emph{reference} lattice $(X,\sqsubseteq)$, we define \emph{fuzzy
48: intervals }to be the fuzzy sets such that their $p$-cuts are crisp closed
49: intervals of $(X,\sqsubseteq)$. We show that: given a complete lattice
50: $(X,\sqsubseteq)$ the collection of its fuzzy intervals is a complete lattice.
51: Furthermore we show that: if $(X,\sqsubseteq)$ is completely distributive then
52: the lattice of its fuzzy intervals is distributive.
53:
54: \vspace{0.5cm}
55:
56: \noindent\textbf{Keywords:} Algebra, Fuzzy Algebras, Fuzzy Lattices.
57:
58: \end{abstract}
59:
60: \section{Introduction}
61:
62: \label{sec01}
63:
64: The following is a small sample of the large literature on fuzzy algebras.
65: Rosenfeld wrote the first paper on \emph{fuzzy groups }\cite{Ros01}; a recent
66: review is \cite{AjGrp01}. \emph{Fuzzy rings }and \emph{fuzzy ideals of rings
67: }are studied in \cite{Yue01,Dixit01,AjRng01,ZahRng01}. Seselja, Tepavcevska
68: and others have presented a far reaching famework of L-fuzzy and P-fuzzy
69: algebras \cite{Ses01,Ses03,Ses04}.
70:
71: \emph{Fuzzy lattices} are a particular type of fuzzy algebras. A fuzzy lattice
72: is a fuzzy set such that its cuts are sublattices of a ``reference lattice''
73: $(X,\sqsubseteq)$. Relatively little has been published on fuzzy lattices.
74: Yuan and Wu introduced the concept \cite{Yuan1} and Ajmal studied it in
75: greater detail \cite{AjLat01}. Swamy and Raju \cite{Swamy1} and, more
76: recently, Tepavcevska and Trajkovski \cite{Tep01} studied \emph{L-fuzzy
77: lattices}.\footnote{Two additional senses of the term ``fuzzy lattice'' should
78: also be mentioned. Kaburlasos and Petridis use \emph{fuzzy inclusion measures}
79: \cite{VGK03,VGK01,VGK02} to introduce a concept of ``fuzzy lattice'' which is
80: different from the one used in the previously mentioned works; however there
81: is an interesting connection between the two approaches through the concept of
82: \emph{fuzzy orders}. In addition, \cite{Wang01,Wang03,Wang02} and many others
83: use the term ``fuzzy lattice'' to denote a quite different mathematical
84: concept, namely a completely distributive lattice with an order reversing
85: involution.}.
86:
87: In this note we introduce \emph{fuzzy intervals }within the context of fuzzy
88: lattices. I.e. a fuzzy interval is defined to be a fuzzy set such that its
89: cuts are closed intervals of a reference lattice $(X,\sqsubseteq)$. It appears
90: that fuzzy intervals (in this lattice theoretic sense) have not been studied
91: previously. A special case which has been extensively studied is that of fuzzy
92: intervals with the reference lattice $(X,\sqsubseteq)$ being a set of
93: \emph{real numbers }\cite{Nguyen}. Some connections between this special case
94: and the more general case studied here will be discussed briefly in Section
95: \ref{sec05}..
96:
97: As mentioned, our study of fuzzy intervals is lattice theoretic. We establish
98: some basic properties of fuzzy intervals and we show the following: given a
99: complete lattice $(X,\sqsubseteq)$, the collection of its fuzzy intervals is a
100: complete lattice; if $(X,\sqsubseteq)$ is \emph{completely distributive} then
101: the lattice of its fuzzy intervals is distributive.
102:
103: \section{Preliminaries}
104:
105: \label{sec02}
106:
107: In what follows, the closed unit interval is denoted by $L\doteq
108: \lbrack0,1]\subseteq R$. The usual order of real numbers is denoted by $\leq$;
109: the maximum (resp. minimum) of $x,y$ is denoted by $x\vee y\ $\ (resp.
110: $x\wedge y $). Given a set $P\subseteq L$, $\vee P$ (resp. $\wedge P$) denotes
111: the supremum (resp. the infimum)\ of $P$. $(L,\leq,\vee,\wedge)$ is a totally
112: ordered set.
113:
114: The \emph{reference lattice }is denoted by $(X,\sqsubseteq,\sqcup,\sqcap)$ and
115: it is assumed to be complete. Hence, for every $Y\subseteq X$ the elements
116: $\sqcap Y$, $\sqcup Y$ exist; in particular, there exist $\sqcap X$ (the
117: minimum element of $X$) and $\sqcup X$ (the maximum element of $X$), hence we
118: can write $X=[\sqcap X,\sqcup X]$.
119:
120: \begin{definition}
121: \label{cnt0201}A \emph{fuzzy set} is a function $M:X\rightarrow L$. The
122: collection of all fuzzy sets (from $X$ to $L$) will be denoted by
123: $\mathbf{F}(X,L)$ or simply by $\mathbf{F}.$
124: \end{definition}
125:
126: In a standard manner, we introduce an order on $\mathbf{F}$ using the
127: ``pointwise'' order of $(L,\leq,\vee,\wedge)$. The symbols $\leq,\vee,\wedge$
128: will be used without danger of confusion.
129:
130: \begin{definition}
131: \label{cnt0202}For $M,N\in\mathbf{F}$ we write $M\leq N$ iff for all $x\in X$
132: we have: $M(x)\leq N(x)$.
133: \end{definition}
134:
135: \begin{definition}
136: \label{cnt0203}For $M,N\in\mathbf{F}$: we define the fuzzy set $M\vee N$ by:
137: $(M\vee N)(x)\doteq M(x)\vee N(x)$; we define the fuzzy set $M\wedge N$ by:
138: $(M\wedge N)(x)\doteq M(x)\wedge N(x)$.
139: \end{definition}
140:
141: It is well known \cite{Nguyen} that $\leq$ is an order on $\mathbf{F}$ and
142: that $(\mathbf{F},\leq,\vee,\wedge)\ $is a complete and distributive lattice
143: with $\sup(M,N)=M\vee N$, $\inf(M,N)=M\wedge N$.
144:
145: \begin{definition}
146: \label{cnt0204}Given a fuzzy set $M:X\rightarrow L$, the $p$-\emph{cut }of $M
147: $ is denoted by $M_{p}$ and defined by $M_{p}\doteq\{x:M(x)\geq p\}.$
148: \end{definition}
149:
150: We will need some properties of $p$-cuts, summarized in the following
151: propositions. Their proofs can be found in \cite{Nguyen}.
152:
153: \begin{proposition}
154: \label{cnt0205}Take any $M\in\mathbf{F}$ with $p$-cuts $\left\{
155: M_{p}\right\} _{p\in L}$ and $N\in\mathbf{F}$ with $p$-cuts $\left\{
156: N_{p}\right\} _{p\in L}$. Then $M=N$ iff $\ $for all $p\in L$ we have
157: $M_{p}=N_{p}$.
158: \end{proposition}
159:
160: \begin{proposition}
161: \label{cnt0206}Take any $M\in\mathbf{F}$ with $p$-cuts $\left\{
162: M_{p}\right\} _{p\in L}$. Then we have the following.
163:
164: (i)\ For all $p,q\in L$ we have: $p\leq q\Rightarrow M_{q}\subseteq M_{p}$.
165:
166: (ii)\ For all $P\subseteq L$ we have: $\cap_{p\in P}M_{p}=M_{\vee P}$.
167:
168: (iii) $M_{0}=X$.
169: \end{proposition}
170:
171: \begin{proposition}
172: \label{cnt0207}Consider a family of sets $\{\widetilde{M}_{p}\}_{p\in L}$
173: which satisfy the following.
174:
175: (i)$\;$For all $p,q\in L$ we have: $p\leq q\Rightarrow\widetilde{M}%
176: _{q}\subseteq\widetilde{M}_{p}$.
177:
178: (ii)\ For all $P\subseteq L$ we have: $\cap_{p\in P}\widetilde{M}%
179: _{p}=\widetilde{M}_{\vee P}$.
180:
181: (iii) $\widetilde{M}_{0}=X$.
182:
183: Define the fuzzy set $M(x)=\vee\{p:x\in\widetilde{M}_{p}\}$. Then for all
184: $p\in L$ we have $M_{p}=\widetilde{M}_{p}.$
185: \end{proposition}
186:
187: We will also need some well-known properties of (crisp)\ closed intervals in a lattice.
188:
189: \begin{definition}
190: \label{cnt0208}Given $x_{1},x_{2}\in X$, with $x_{1}\sqsubseteq x_{2}$, the
191: \emph{closed interval} $[x_{1},x_{2}]$ is defined by $[x_{1},x_{2}]\doteq$
192: $\{z:x_{1}\sqsubseteq z\sqsubseteq x_{2}\}.$
193: \end{definition}
194:
195: We consider the empty set $\emptyset$ to be a closed interval, the so called
196: \emph{empty} \emph{interval}. This can also be denoted as $[x_{1},x_{2}]$ with
197: any $x_{1},x_{2}$ such that $x_{1}\not \sqsubseteq x_{2}$. Denote by
198: $\mathbf{I}\ $\ the collection of (crisp) closed intervals of $X$ (including
199: the empty interval). The structure $(\mathbf{I},\subseteq)$ is an ordered set.
200: In fact it is a lattice, as the following propositions show (proofs are
201: omitted for brevity; they follow from the fact that being a closed interval is
202: a \emph{closure property} on $(\mathbf{I},\subseteq)$ \cite{Birkhoff}).
203:
204: \begin{proposition}
205: \label{cnt0209}Given any nonempty interval $A=[a_{1},a_{2}]$ $\subseteq X$, we
206: have $a_{1}=\sqcap A$, $a_{2}=\sqcup A$.
207: \end{proposition}
208:
209: \begin{proposition}
210: \label{cnt0210}Given any family of closed intervals $\mathbf{J\subseteq I}$
211: the set $\cap_{\lbrack a_{1},a_{2}]\in\mathbf{J}}[a_{1},a_{2}]$ is a closed
212: interval; more specifically, we have
213: \[
214: \cap_{\lbrack a_{1},a_{2}]\in\mathbf{J}}[a_{1},a_{2}]=[\sqcup_{\lbrack
215: a_{1},a_{2}]\in\mathbf{J}}a_{1},\sqcap_{\lbrack a_{1},a_{2}]\in\mathbf{J}%
216: }a_{2}]
217: \]
218: and this is the largest closed interval contained in every member of
219: $\mathbf{J}$.
220: \end{proposition}
221:
222: \begin{definition}
223: \label{cnt0211}Given $A,B\in\mathbf{I}$, define $\mathbf{S}(A,B)\mathbf{\doteq
224: }$ $\mathbf{\{}C$: $C\in\mathbf{I}$, $A\subseteq C,B\subseteq C\mathbf{\}}$.
225: Then we define
226: \[
227: A\overset{.}{\cup}B\doteq\cap_{C\in\mathbf{S}(A,B)}C.
228: \]
229: \end{definition}
230:
231: \begin{proposition}
232: \label{cnt0212}The structure $(\mathbf{I},\subseteq,\overset{.}{\cup},\cap)$
233: is a lattice with respect to the $\subseteq$ order (i.e. set theoretic
234: inclusion). Given any intervals $A=[a_{1},a_{2}]$ $\in\mathbf{I}$,
235: $B=[b_{1},b_{2}]\in\mathbf{I}$, $\sup(A,B)$ = $A\overset{.}{\cup}B$ =
236: $[a_{1}\sqcap b_{1},a_{2}\sqcup b_{2}]$, \ $\inf(A,B)$= $A\cap$ $B$ =
237: $[a_{1}\sqcup b_{1}$ , $a_{2}\sqcap b_{2}]$.
238: \end{proposition}
239:
240: \textbf{Remark}. In other words, given any intervals $A=[a_{1},a_{2}]$,
241: $B=[b_{1},b_{2}]$, $[a_{1}\sqcap b_{1},a_{2}\sqcup b_{2}]$ is the smallest
242: closed interval which contains both $A$ and $B\ $and [$a_{1}\sqcup b_{1}$ ,
243: $a_{2}\sqcap b_{2}$] is the largest closed interval contained in both $A$ and
244: $B$.
245:
246: We define \emph{fuzzy sublattices} and \emph{fuzzy convex sublattices} in
247: terms of their $p$-cuts; this is different from, but equivalent to Ajmal's
248: approach \cite{AjLat01}.
249:
250: \begin{definition}
251: \label{cnt0213}We say $M:X\rightarrow L$ is a \emph{fuzzy sublattice} of
252: $(X,\sqsubseteq)$ iff $\forall p\in L$ the set$\ M_{p}$ is a sublattice of
253: $(X,\sqsubseteq).$
254: \end{definition}
255:
256: \begin{definition}
257: \label{cnt0214}We say $M:X\rightarrow L$ is a \emph{fuzzy convex sublattice}
258: of $(X,\sqsubseteq)$ iff $\forall p\in L$ the set$\ M_{p}$ is a convex
259: sublattice of $(X,\sqsubseteq)$; (i.e. $\forall p\in L,\forall x,y\in M_{p}$
260: we have $[x\sqcap y,x\sqcup y]\subseteq M_{p}$).
261: \end{definition}
262:
263: \begin{proposition}
264: \label{cnt0215}$M:X\rightarrow L$ is a fuzzy sublattice of $(X,\sqsubseteq)$
265: iff
266: \[
267: \forall x,y\in X:\quad M(x\sqcap y)\wedge M(x\sqcup y)\geq M(x)\wedge M(y).
268: \]
269: \end{proposition}
270:
271: \begin{proof}
272: See \cite{Tep01}.
273: \end{proof}
274:
275: \begin{proposition}
276: \label{cnt0216}Let $M:X\rightarrow L$ be a fuzzy sublattice of $(X,\sqsubseteq
277: )$. It is a fuzzy convex sublattice of $(X,\sqsubseteq)$ iff
278: \begin{equation}
279: \forall x,y\in X,\forall z\in\lbrack x\sqcap y,x\sqcup y]:\quad M(z)\geq
280: M(x\sqcap y)\wedge M(x\sqcup y)=M(x)\wedge M(y). \label{Eq02}%
281: \end{equation}
282: \end{proposition}
283:
284: \begin{proof}
285: (i)\ Assume $M$ is a fuzzy convex sublattice. Choose any $x,y\in X$. Set
286: $p_{1}=M(x\sqcap y)$, $p_{2}=M(x\sqcup y)$; then $x\sqcap y$, $x\sqcup y\in$
287: $M_{p_{1}\wedge p_{2}}$. Take any $z\in\lbrack x\sqcap y,x\sqcup y]$. Since
288: $M$ is a fuzzy convex sublattice: $z\in M_{p_{1}\wedge p_{2}}\Rightarrow$
289: $M(z)\geq p_{1}\wedge p_{2}$ = $M(x\sqcap y)\wedge M(x\sqcup y)$. Since
290: $x,y\in$\ $[x\sqcap y,x\sqcup y]$ we have $M(x)\geq M(x\sqcap y)\wedge
291: M(x\sqcup y)$, $M(y)\geq M(x\sqcap y)\wedge M(x\sqcup y)$; and so $M(x)\wedge
292: M(y)\geq M(x\sqcap y)\wedge M(x\sqcup y)$. On the other hand, since $M$ is a
293: fuzzy sublattice, from Proposition \ref{cnt0215} we have $M(x\sqcap y)\wedge
294: M(x\sqcup y)\geq$ $M(x)\wedge M(y)$. Hence $M(x\sqcap y)\wedge M(x\sqcup
295: y)\ \ $= $M(x)\wedge M(y)$.
296:
297: (ii)\ Conversely, assume (\ref{Eq02}) holds. Take any $p\in L$. If $M_{p}$ is
298: empty, \ then it is a convex sublattice. If $M_{p}$ is not empty, take any
299: $x,y\in M_{p}$. Set $p_{1}=M(x)$, $p_{2}=M(y)$. We have $x\in M_{p}%
300: \Rightarrow$ $p_{1}=M(x)\geq p$, $y\in M_{p}\Rightarrow$ $p_{2}=M(y)\geq p$.
301: From (\ref{Eq02}) we have $M(x\sqcap y)\geq M(x)\wedge M(y)=p_{1}\wedge
302: p_{2}\geq p\Rightarrow$ $x\sqcap y\in M_{p}$. Similarly $x\sqcup y\in M_{p}$
303: \ and so $M_{p}$ is a sublattice. Set $q_{1}=M(x\sqcap y)$, $q_{2}=M(x\sqcup
304: y)$. Now take any $z\in\lbrack x\sqcap y,x\sqcup y]$. From (\ref{Eq02}) we
305: have $M(z)\geq q_{1}\wedge q_{2}=p_{1}\wedge p_{2}\geq p\Rightarrow$ $z\in
306: M_{p}$. Hence $M_{p}$ is a convex sublattice for all $p\in L$, i.e. $M\ $ is a
307: fuzzy convex sublattice.
308: \end{proof}
309:
310: \section{The Lattice of Fuzzy Intervals}
311:
312: \label{sec03}
313:
314: We now introduce \emph{fuzzy intervals}.
315:
316: \begin{definition}
317: \label{cnt0301}We say $M:X\rightarrow L$ is a \emph{fuzzy interval }\ of
318: $(X,\sqsubseteq)$ iff
319: \[
320: \forall p\in L:\quad M_{p}\text{ is a closed interval of }(X,\leq).
321: \]
322: The collection all fuzzy intervals will be denoted by $\widetilde{\mathbf{I}%
323: }(X,L)$ or simply by $\widetilde{\mathbf{I}}$.
324: \end{definition}
325:
326: The following proposition will be often used in the sequel. It states that an
327: arbitrary intersection of fuzzy intervals yields a fuzzy interval.
328:
329: \begin{proposition}
330: \label{cnt0302}For all $\widetilde{\mathbf{J}}\subseteq\widetilde{\mathbf{I}}$
331: we have: $\wedge_{M\in\widetilde{\mathbf{J}}}M\in\widetilde{\mathbf{I}} $
332: \end{proposition}
333:
334: \begin{proof}
335: Choose any $\widetilde{\mathbf{J}}\subseteq\widetilde{\mathbf{I}}%
336: \subseteq\mathbf{F}$. The fuzzy set $\wedge_{M\in\widetilde{\mathbf{J}}}M$ is
337: well defined, in view of the fact that $(\mathbf{F},\leq,\vee,\wedge)$ is a
338: complete lattice. Choose any $p\in L$. It is easy to show that $(\wedge
339: _{M\in\widetilde{\mathbf{J}}}M)_{p}$ = $\cap_{M\in\widetilde{\mathbf{J}}}%
340: M_{p}$. Then for every $M\in\widetilde{\mathbf{J}}$ , the cut $M_{p}$ will be
341: a closed interval (perhaps the empty interval). From Proposition
342: \ref{cnt0210}, an arbitrary intersection of closed intervals yields a closed
343: interval. Hence, for every $p\in L$ the set $(\wedge_{M\in\widetilde
344: {\mathbf{J}}}M)_{p}$ is a closed interval, i.e. $\wedge_{M\in\widetilde
345: {\mathbf{J}}}M$ is a fuzzy interval.
346: \end{proof}
347:
348: Since $\widetilde{\mathbf{I}}\subseteq\mathbf{F}$, it follows that
349: $(\widetilde{\mathbf{I}},\leq)\ $is an ordered set. We now establish (using
350: Proposition \ref{cnt0302}) that $(\widetilde{\mathbf{I}},\leq)$ is a lattice.
351:
352: \begin{definition}
353: \label{cnt0303}For all $M,N\in\widetilde{\mathbf{I}}$ we define $M\overset
354: {.}{\vee}N$ as follows. We define $\widetilde{\mathbf{S}}(M,N)\doteq$
355: $\{A:A\in\widetilde{\mathbf{I}}$, $M\leq A,N\leq A\}$ and then define
356: \[
357: M\overset{.}{\vee}N\doteq\wedge_{A\in\widetilde{\mathbf{S}}(M,N)}A.
358: \]
359: \end{definition}
360:
361: \begin{proposition}
362: \label{cnt0304}$(\widetilde{\mathbf{I}},\leq,\overset{.}{\vee},\wedge)\ $is a
363: complete lattice.
364: \end{proposition}
365:
366: \begin{proof}
367: (i)$\;M\wedge N$ is the infimum in $\mathbf{F}$ of $M$ and $N$. From
368: Proposition \ref{cnt0302} we have $M\wedge N\in\widetilde{\mathbf{I}}$, hence
369: $M\wedge N$ is also the infimum of $M$ and $N$ in $\widetilde{\mathbf{I}}$.
370:
371: (ii)\ For all $A\in\widetilde{\mathbf{S}}(M,N)$ we have $M\leq A$ and so
372: $M\leq\wedge_{A\in\widetilde{\mathbf{S}}(M,N)}A=M\overset{.}{\vee}N$;
373: similarly $N\leq M\overset{.}{\vee}N$. Furthermore, if there is some
374: $B\in\widetilde{\mathbf{I}}$ such that $M\leq B$, $N\leq B$, then
375: $B\in\widetilde{\mathbf{S}}(M,N)$. Hence $M\overset{.}{\vee}N$ = $\wedge
376: _{A\in\widetilde{\mathbf{S}}(M,N)}A\leq B$. \ Finally, since $\widetilde
377: {\mathbf{S}}(M,N)$ $\subseteq\widetilde{\mathbf{I}}$, we have $M\overset
378: {.}{\vee}N$ = $\wedge_{A\in\widetilde{\mathbf{S}}(M,N)}A\in\widetilde
379: {\mathbf{I}}$. Hence $M\overset{.}{\vee}N$ is the supremum in $\widetilde
380: {\mathbf{I}}$ of $M$ and $N$.\noindent
381:
382: (iii) To establish completeness of $(\widetilde{\mathbf{I}},\leq,\overset
383: {.}{\vee},\wedge)\ $we must show that any $\widetilde{\mathbf{J}}%
384: \subseteq\widetilde{\mathbf{I}}$ has an infimum and a supremum in
385: $\widetilde{\mathbf{I}}$. We have already remarked (Proposition \ref{cnt0302})
386: that, for any $\widetilde{\mathbf{J}}\subseteq\widetilde{\mathbf{I}}$, the set
387: $\wedge_{M\in\widetilde{\mathbf{J}}}M$ is a well defined fuzzy interval. Since
388: $\wedge\widetilde{\mathbf{J}}$ $\ $ = $\wedge_{M\in\widetilde{\mathbf{J}}}M $
389: is the infimum of $\widetilde{\mathbf{J}}$ in $\mathbf{F}$, it will also be
390: the infimum of $\widetilde{\mathbf{J}}$ in $\widetilde{\mathbf{I}}%
391: \subseteq\mathbf{F}$. Regarding the supremum, we must define appropriately
392: $\overset{.}{\vee}\widetilde{\mathbf{J}}$. Define a set $\widetilde
393: {\mathbf{S}}(\widetilde{\mathbf{J}})=\{A\in\widetilde{\mathbf{I}}:$ $\forall
394: M\in\widetilde{\mathbf{J}}$ we have $\ M\leq A\}$. Define $\overset{.}{\vee
395: }\widetilde{\mathbf{J}}\doteq\wedge_{A\in\widetilde{\mathbf{S}}(\widetilde
396: {\mathbf{J}})}A$. Then $\overset{.}{\vee}\widetilde{\mathbf{J}}\in
397: \widetilde{\mathbf{I}}$ (as an intersection of fuzzy intervals), and it is
398: easy to show that: $\forall M\in\widetilde{\mathbf{J}}$ we have $M\leq
399: \overset{.}{\vee}\widetilde{\mathbf{J}}$, $\forall A\in\widetilde{\mathbf{S}%
400: }(\widetilde{\mathbf{J}})$ we have $\overset{.}{\vee}\widetilde{\mathbf{J}%
401: }\leq A$. Hence $\overset{.}{\vee}\widetilde{\mathbf{J}}$ is the supremum of
402: $\widetilde{\mathbf{J}}$ and completeness has been established.
403: \end{proof}
404:
405: The following propositions establish some properties of fuzzy intervals.
406:
407: \begin{definition}
408: \label{cnt0305}For every fuzzy set $M$ we define $L_{M}\doteq\{p:M_{p}%
409: \neq\emptyset\}$.
410: \end{definition}
411:
412: \begin{proposition}
413: \label{cnt0306}(i)\ Let $M$ be a fuzzy convex sublattice. If we have
414: \begin{equation}
415: \forall p\in L_{M}:M(\sqcap M_{p})\geq\wedge_{x\in M_{p}}M(x),\quad M(\sqcup
416: M_{p})\geq\wedge_{x\in M_{p}}M(x), \label{Eq31}%
417: \end{equation}
418: then $M$ is a fuzzy interval.
419:
420: (ii) If $M$ is a fuzzy interval, then it is a fuzzy convex sublattice and we
421: have
422: \[
423: \forall p\in L_{M}:M(\sqcap M_{p})\geq\wedge_{x\in M_{p}}M(x),\quad M(\sqcup
424: M_{p})\geq\wedge_{x\in M_{p}}M(x).
425: \]
426: \end{proposition}
427:
428: \begin{proof}
429: (i) Assume (\ref{Eq31}) holds. Choose any $p\in L_{M}$. Now, by completenes of
430: $(X,\sqsubseteq)$, $\sqcap M_{p}$ and $\sqcup M_{p}$ exist. Clearly
431: $M_{p}\subseteq\lbrack\sqcap M_{p},\sqcup M_{p}]$. On the other hand, from
432: (\ref{Eq31}), $M(\sqcap M_{p})\geq\wedge_{x\in M_{p}}M(x)\geq p\Rightarrow$
433: $\sqcap M_{p}\in M_{p}$, i.e. $M_{p}$ contains its infimum. Similarly
434: $M(\sqcup M_{p})\geq\wedge_{x\in M_{p}}M(x)\geq p\Rightarrow$ $\sqcup M_{p}\in
435: M_{p}$. Since $M_{p}$ is a convex sublattice and $\sqcap M_{p},\sqcup M_{p}\in
436: M_{p}$, it follows that $[\sqcap M_{p},\sqcup M_{p}]\subseteq M_{p}$. Hence
437: for all $p\in L_{M}$ we have that $M_{p}=[\sqcap M_{p},\sqcup M_{p}]$.
438: Further, for all $p\in L-L_{M}$, $M_{p}$ is the empty set, which is considered
439: a closed interval. Hence for all $p\in L$ the set $M_{p}$ is a closed
440: interval, i.e. $M$ is a fuzzy interval.
441:
442: (ii) If $M$ is a fuzzy interval then for all $p\in L_{M}$ we have $M_{p}$ =
443: [$\sqcap M_{p}$, $\sqcup M_{p}$], which is a closed interval and \emph{a
444: fortiori} a convex sublattice. Hence$\;M$ is a fuzzy convex sublattice.
445: Furthermore, $M_{p}$ = [$\sqcap M_{p}$, $\sqcup M_{p}$]$\Rightarrow$ $\sqcap
446: M_{p}\in M_{p}$ $\Rightarrow$ $M(\sqcap M_{p})$ $\geq$ $\wedge_{x\in M_{p}%
447: }M(x)$. Similarly, $\sqcup M_{p}\in M_{p}$ $\Rightarrow$ $M(\sqcup M_{p})$
448: $\geq$ $\wedge_{x\in M_{p}}M(x)$ .
449: \end{proof}
450:
451: \begin{corollary}
452: \label{cnt0307}If $M$ is a fuzzy interval, then $\forall p\in L_{M}$ we have
453: $M(\sqcap M_{p})\wedge M(\sqcup M_{p})$ = $\wedge_{x\in M_{p}}M(x).$
454: \end{corollary}
455:
456: \begin{corollary}
457: \label{cnt0308}Let $X$ be finite. Then every fuzzy convex sublattice is a
458: fuzzy interval and conversely.
459: \end{corollary}
460:
461: \begin{proposition}
462: \label{cnt0309}If $M$ is a fuzzy interval, then $\forall p\in L_{M}$ we have
463: $M_{p}$ = $M_{p_{1}\wedge p_{2}}$, where $p_{1}=M(\sqcap M_{p})$,
464: $p_{2}=M(\sqcup M_{p}).$
465: \end{proposition}
466:
467: \begin{proof}
468: Choose any $p\in L_{M}$. Since $M$ is a fuzzy interval, we have $M_{p}=[\sqcap
469: M_{p},\sqcup M_{p}].\ $Set $p_{1}=M(\sqcap M_{p})\geq p$,\ $p_{2}=M(\sqcup
470: M_{p})\geq p$. Then $M(\sqcap M_{p})=p_{1}\geq p_{1}\wedge p_{2}$ and so
471: $\sqcap M_{p}\in M_{p_{1}\wedge p_{2}}$. Similarly $\sqcup M_{p}\in
472: M_{p_{1}\wedge p_{2}}$. Since $M$ is a fuzzy interval (and so a fuzzy convex
473: sublattice)\ it follows that $[\sqcap M_{p},\sqcup M_{p}]\subseteq
474: M_{p_{1}\wedge p_{2}}$. On the other hand $p_{1}\wedge p_{2}\geq p\Rightarrow$
475: $M_{p_{1}\wedge p_{2}}\subseteq M_{p}$ = $[\sqcap M_{p},\sqcup M_{p}]$. Hence
476: $M_{p_{1}\wedge p_{2}}$ = $M_{p}$.
477: \end{proof}
478:
479: \section{Distributivity}
480:
481: \label{sec04}
482:
483: In all of this section we assume $(X,\sqsubseteq,\sqcup,\sqcap)$ to be
484: \emph{completely distributive} according to the following definition.
485:
486: \begin{definition}
487: \label{cnt0401}The lattice $(X,\sqsubseteq,\sqcup,\sqcap)$ is said to be
488: \emph{completely distributive}, iff for every set $Y\subseteq X$ we have
489: $x\sqcup(\sqcap_{y\in Y}y)=\sqcap_{y\in Y}(x\sqcup y),\qquad x\sqcap
490: (\sqcup_{y\in Y}y)=\sqcup_{y\in Y}(x\sqcap y).$
491: \end{definition}
492:
493: Let $M,N$ be fuzzy intervals. Our first task is to establish some properties
494: of the cuts $\left( M\wedge N\right) _{p}\ $\ and $(M\overset{.}{\vee}N)_{p}
495: $. From Proposition \ref{cnt0304} we see that $M\wedge N\ $\ and $M\overset
496: {.}{\vee}N$ are fuzzy intervals; hence $\forall p\in L$ the cuts $\left(
497: M\wedge N\right) _{p}\ $\ and $(M\overset{.}{\vee}N)_{p}$ are (crisp) closed intervals.
498:
499: \begin{definition}
500: \label{cnt0402}For all $M,N\in\widetilde{\mathbf{I}}$ and for all $p\in L$ we
501: define $C_{p}(M,N)$ $=M_{p}\cap N_{p}$.
502: \end{definition}
503:
504: \begin{proposition}
505: \label{cnt0403}For all $M,N\in\widetilde{\mathbf{I}}$ and for all $p\in L$ we
506: have: $(M\wedge N)_{p}=C_{p}(M,N\mathbf{)}$.
507: \end{proposition}
508:
509: \begin{proof}
510: Take any $M,N\in\widetilde{\mathbf{I}}$, any $p\in L$. We have $x\in(M\wedge
511: N)_{p}$ $\Leftrightarrow$ $\left( M\wedge N\right) (x)\geq p$
512: $\Leftrightarrow$ $M(x)\wedge N(x)\geq p$ $\Leftrightarrow$ $\left( M(x)\geq
513: p\text{ and }N(x)\geq p\right) $ $\Leftrightarrow$ $\left( x\in M_{p}\text{
514: and }x\in N_{p}\right) $ $\Leftrightarrow$ $x\in M_{p}\cap N_{p}%
515: =C_{p}(M,N\mathbf{)}$.
516: \end{proof}
517:
518: \begin{proposition}
519: \label{cnt0404}Take any $M,N\in\widetilde{\mathbf{I}}$. We have:
520:
521: (i)\ $\forall p,q\in L$ :\ $p\leq q\Rightarrow$ $C_{q}(M,N)$ $\subseteq
522: C_{p}(M,N)$,
523:
524: (ii) $\forall P\subseteq L$: $\cap_{p\in P}C_{p}(M,N)$ = $C_{\vee P}(M,N)$.
525:
526: (iii) $C_{0}(M,N)=X$.
527: \end{proposition}
528:
529: \begin{proof}
530: These properties follow from the fact that for all $p\in L$ we have
531: $C_{p}(M,N\mathbf{)}$ = $(M\wedge N)_{p}$, i.e. the family $\{C_{p}%
532: (M,N\mathbf{)\}}_{p\in L}$ is a family of cuts.
533: \end{proof}
534:
535: Hence we have characterized the cuts of $M\wedge N$ in terms of the cuts of
536: $M$ and $N$. We will now do the same for the cuts of $M\overset{.}{\vee}N$.
537: However, before proceeding we need some auxiliary definitions and propositions.
538:
539: \begin{definition}
540: \label{cnt0405}For every $M\in\widetilde{\mathbf{I}}$, we define the functions
541: $\underline{M}:L\rightarrow X$, $\overline{M}:L\rightarrow X$ as follows. For
542: $p\in L_{M}$, $\underline{M}(p)\doteq\sqcap M_{p}$, $\overline{M}%
543: (p)\doteq\sqcup M_{p}$; for $p\in L-L_{M}$, $\underline{M}(p)\doteq\sqcup X$,
544: $\overline{M}(p)\doteq\sqcap X$.
545: \end{definition}
546:
547: \textbf{Remark}. Hence we can write $M_{p}=[\underline{M}(p),\overline{M}(p)]
548: $ for \emph{every }$p\in L$. Because: if $p\in L_{M}$, then $M_{p}$ = [$\sqcap
549: M_{p},\sqcup M_{p}$] = $[\underline{M}(p),\overline{M}(p)]$; if $p\in L-L_{M}%
550: $, then $M_{p}=\emptyset$ = [$\sqcup X,\sqcap X$] = $[\underline
551: {M}(p),\overline{M}(p)]$.
552:
553: \begin{proposition}
554: \label{cnt0406}Take any $M\in\widetilde{\mathbf{I}}$ and for all $p\in L$ set
555: $M_{p}=[\underline{M}(p),\overline{M}(p)]$. Then
556:
557: (i) $\forall p,q\in L:p\leq q\Rightarrow\left( \underline{M}(p)\sqsubseteq
558: \underline{M}(q),\overline{M}(p)\sqsupseteq\overline{M}(q)\right) $ .
559:
560: (ii)\ $\forall P\subseteq L:\sqcup_{p\in P}\underline{M}(p)=\underline{M}(\vee
561: P)$, $\sqcap_{p\in P}\overline{M}(p)=\overline{M}(\vee P)$.
562: \end{proposition}
563:
564: \begin{proof}
565: (i)\ Since $\left\{ M_{p}\right\} _{p\in P}$ are cuts, from
566: Prop.\ref{cnt0206}.(i) we have: $p\leq q\Rightarrow M_{q}\subseteq
567: M_{p}\Rightarrow$ [$\underline{M}(q)$ , $\overline{M}(q)$] $\subseteq$
568: [$\underline{M}(p)$ , $\overline{M}(p)$]$\Rightarrow$ ($\underline{M}%
569: (p)\leq\underline{M}(q)$ , $\overline{M}(p)\geq\overline{M}(q)$). Note in
570: particular that: if $q\notin L_{M}$, then $\underline{M}(p)\sqsubseteq
571: \underline{M}(q)=\sqcup X$ and $\overline{M}(p)\sqsupseteq\overline
572: {M}(q)=\sqcap X$.
573:
574: (ii) Since $\left\{ M_{p}\right\} _{p\in P}$ are cuts, from
575: Prop.\ref{cnt0206}.(ii) we have:\ $\cap_{p\in P}M_{p}=M_{\vee P}$. But
576: $M_{\vee P}$\ = [\underline{$M$}$(\vee P)$ , $\overline{M}(\vee P)$] and
577: (Proposition \ref{cnt0210})\ $\cap_{p\in P}M_{p}$= [$\sqcup_{p\in P}%
578: \underline{M}(p)$ , $\sqcap_{p\in P}\overline{M}(p)$] which yields the
579: required result. Note in particular that:\ if there exists some $q\in P$ such
580: that $q\in L-L_{M}$, then $M_{q}=\emptyset$, \ $\cap_{p\in P}M_{p}=\emptyset$,
581: and $M_{\vee P}=\emptyset$ = $[\underline{M}(\vee P),\overline{M}(\vee P)]$
582: with $\underline{M}(\vee P)$ = $\sqcup X$, $\overline{M}(\vee P)$ = $\sqcap
583: X$. Also, in this case $\underline{M}(q)=\sqcup X$, $\sqcup_{p\in P}%
584: \underline{M}(p)=\sqcup X$, $\overline{M}(q)=\sqcap X$, $\sqcap_{p\in
585: P}\underline{M}(p)=\sqcap X$.
586: \end{proof}
587:
588: \begin{proposition}
589: \label{cnt0407}(i) Take any $P\subseteq L$ and any functions $F:L\rightarrow
590: X$, $G:L\rightarrow X$ which satisfy
591: \begin{align*}
592: p & \leq q\Rightarrow F(p)\sqsubseteq F(q),\qquad\sqcup_{p\in P}F(p)=F(\vee
593: P),\\
594: p & \leq q\Rightarrow G(p)\sqsubseteq G(q),\qquad\sqcup_{p\in P}G(p)=G(\vee
595: P).
596: \end{align*}
597: Then $\sqcup_{p\in P}\left( F(p)\sqcap G(p)\right) =F(\vee P)\sqcap G(\vee
598: P)$.
599:
600: (ii) Take any $P\subseteq L$ and any functions $F:L\rightarrow X$,
601: $G:L\rightarrow X$ which satisfy
602: \begin{align*}
603: p & \leq q\Rightarrow F(p)\sqsupseteq F(q),\qquad\sqcap_{p\in P}F(p)=F(\vee
604: P),\\
605: p & \leq q\Rightarrow G(p)\sqsupseteq G(q),\qquad\sqcap_{p\in P}G(p)=G(\vee
606: P).
607: \end{align*}
608: Then $\sqcap_{p\in P}\left( F(p)\sqcup G(p)\right) =F(\vee P)\sqcup G(\vee
609: P)$.
610: \end{proposition}
611:
612: \begin{proof}
613: For (i), take any $p\in P$. Then $F(p)\sqcap G(p)\sqsubseteq F(p).$ Hence
614: $\sqcup_{p\in P}\left( F(p)\sqcap G(p)\right) \sqsubseteq\sqcup_{p\in
615: P}F(p)$ = $F(\vee P)$. Similarly $\sqcup_{p\in P}\left( F(p)\sqcap
616: G(p)\right) \sqsubseteq\sqcup_{p\in P}G(p)$ = $G(\vee P)$. It follows that
617: \begin{equation}
618: \sqcup_{p\in P}\left( F(p)\sqcap G(p)\right) \sqsubseteq F(\vee P)\sqcap
619: G(\vee P). \label{Eq41}%
620: \end{equation}
621: On the other hand, using complete distributivity, we have $\sqcup_{p\in P,q\in
622: P}\left( F(p)\sqcap G(q)\right) $ = $\sqcup_{p\in P}\left( F(p)\sqcap
623: \left( \sqcup_{q\in P}G(q)\right) \right) $ = $\sqcup_{p\in P}\left(
624: F(p)\sqcap G(\vee P)\right) $ = $\left( \sqcup_{p\in P}F(p)\right) \sqcap
625: G(\vee P)$ = $F(\vee P)\sqcap G(\vee P)$. In short
626: \begin{equation}
627: F(\vee P)\sqcap G(\vee P)=\sqcup_{p\in P,q\in P}\left( F(p)\sqcap
628: G(q)\right) \label{Eq42}%
629: \end{equation}
630: Finally, since $(L,\leq)$ is totally ordered, $P$ is a sublattice of
631: $(L,\leq);$ so for any $p,q\in P$ we have $p\vee q\in P$. Then $\left( p\leq
632: p\vee q\text{, }q\leq p\vee q\right) \Rightarrow$ $F(p)\sqcap G(q)\sqsubseteq
633: $ $F(p\vee q)\sqcap G(p\vee q)$. So $\sqcup_{p\in P,q\in P}\left( F(p)\sqcap
634: G(q)\right) \sqsubseteq$ $\sqcup_{p\in P,q\in P}\left( F(p\vee q)\sqcap
635: G(p\vee q)\right) \sqsubseteq$ $\sqcup_{r\in P}\left( F(r)\sqcap
636: G(r)\right) $. Hence
637: \begin{equation}
638: \sqcup_{p\in P,q\in P}\left( F(p)\sqcap G(q)\right) \sqsubseteq\sqcup_{p\in
639: P}\left( F(p)\sqcap G(p)\right) \label{Eq43}%
640: \end{equation}
641: From (\ref{Eq41}), (\ref{Eq42}), (\ref{Eq43}) follows that $\sqcup_{p\in
642: P}\left( F(p)\sqcap G(p)\right) $ = $F(\vee P)\sqcap G(\vee P)$ and (i)
643: \ has been proved; (ii)\ is proved dually.
644: \end{proof}
645:
646: Now we return to the cuts of $M\overset{.}{\vee}N$.
647:
648: \begin{definition}
649: \label{cnt0408}For all $M,N\in\widetilde{\mathbf{I}}$ and for all $p\in L$ we
650: define $D_{p}(M,N)=M_{p}\overset{.}{\cup}N_{p}$.
651: \end{definition}
652:
653: \begin{proposition}
654: \label{cnt0409}Take any $M,N\in\widetilde{\mathbf{I}}$ . We have
655:
656: (i) $\forall p,q\in L$:\ $p\leq q\Rightarrow$ $D_{q}(M,N)$ $\subseteq
657: D_{p}(M,N)$,
658:
659: (ii)$\;\forall P\subseteq L$ : $\cap_{p\in P}D_{p}(M,N)$ = $D_{\vee P}(M,N)$.
660:
661: (iii) $D_{0}(M,N)=X$.
662: \end{proposition}
663:
664: \begin{proof}
665: (i) Assume $p\leq q$. Then $\left( M_{q}\subseteq M_{p},N_{q}\subseteq
666: N_{p}\right) \Rightarrow$ $M_{q}\overset{.}{\cup}N_{q}\subseteq M_{p}%
667: \overset{.}{\cup}N_{p}\Rightarrow$ $D_{q}(M,N)$ $\subseteq D_{p}(M,N)$.
668:
669: (ii)\ Take any $P\subseteq L$ and any $p\in P$. We have $D_{p}%
670: (M,N)=[\underline{M}(p)\sqcap\underline{N}(p),\overline{M}(p)\sqcup
671: \overline{N}(p)]$, hence
672: \begin{equation}
673: \cap_{p\in P}D_{p}(M,N)=[\sqcup_{p\in P}(\underline{M}(p)\sqcap\underline
674: {N}(p)),\sqcap_{p\in P}(\overline{M}(p)\sqcup\overline{N}(p))]. \label{Eq44}%
675: \end{equation}
676: Also
677: \begin{equation}
678: D_{\vee P}(M,N)=[\underline{M}(\vee P)\sqcap\underline{N}(\vee P),\overline
679: {M}(\vee P)\sqcup\overline{N}(\vee P)]. \label{Eq45}%
680: \end{equation}
681: Use Proposition \ref{cnt0407}.(i) with $F(p)=\underline{M}(p)$ and
682: $G(p)=\underline{N}(p)$. Then
683: \begin{equation}
684: \sqcup_{p\in P}(\underline{M}(p)\sqcap\underline{N}(p))=\underline{M}(\vee
685: P)\sqcap\underline{N}(\vee P). \label{Eq46}%
686: \end{equation}
687: Use Proposition \ref{cnt0407}.(ii) with $F(p)=\overline{M}(p)$ and
688: $G(p)=\overline{N}(p)$. Then
689: \begin{equation}
690: \sqcap_{p\in P}(\overline{M}(p)\sqcup\overline{N}(p))=\overline{M}(\vee
691: P)\sqcup\overline{N}(\vee P). \label{Eq47}%
692: \end{equation}
693: Eqs.(\ref{Eq44}--\ref{Eq47})\ yield the required result.
694:
695: (iii)\ $D_{0}(M,N)=M_{0}\overset{.}{\cup}N_{0}=X\overset{.}{\cup}X=X$.
696: \end{proof}
697:
698: \begin{proposition}
699: \label{cnt0410}For all $M,N\in\widetilde{\mathbf{I}}$ and for all $p\in L$ we
700: have: $(M\overset{.}{\vee}N)_{p}=D_{p}(M,N\mathbf{)}$
701: \end{proposition}
702:
703: \begin{proof}
704: From Proposition \ref{cnt0409} follows that $\left\{ D_{p}(M,N)\right\}
705: _{p\in L}$ is a family of cuts. Hence, if we define a fuzzy set $(M\veebar N)
706: $ by setting
707: \[
708: \forall x\in X:(M\veebar N)(x)\doteq\vee\left\{ p:x\in D_{p}(M,N)\right\}
709: \]
710: then $\forall p\in L$ we will have $(M\veebar N)_{p}=D_{p}(M,N)$ (Proposition
711: \ref{cnt0207}). From this also follows that $(M\veebar N)$ is a fuzzy interval
712: (since $\forall p\in L$ we have $(M\veebar N)_{p}$ = $D_{p}(M,N)$ =
713: $M_{p}\overset{.}{\cup}N_{p}$). Now choose any $p\in L$ ; we will show that
714: $(M\overset{.}{\vee}N)_{p}=(M\veebar N)_{p}$.
715:
716: First, $(M\overset{.}{\vee}N)_{p}$ is a (crisp) closed interval. Also, $x\in
717: M_{p}\Rightarrow$ $(M\overset{.}{\vee}N)(x)\geq M(x)\geq p\ \Rightarrow$
718: $x\in(M\overset{.}{\vee}N)_{p}$. So $M_{p}\subseteq(M\overset{.}{\vee}N)_{p}$.
719: Similarly $N_{p}\subseteq(M\overset{.}{\vee}N)_{p}$. Hence $(M\overset{.}%
720: {\vee}N)_{p}\in\mathbf{S}(M_{p},N_{p})$ which implies that $(M\veebar
721: N)_{p}=D_{p}(M,N)$ = $M_{p}\overset{.}{\cup}N_{p}$ = $\cap_{A\in
722: \mathbf{S}(M_{p},N_{p})}A\subseteq$ $(M\overset{.}{\vee}N)_{p}$.
723:
724: Second, choose any $x\in X$ and set $p=M(x)$. Then $x\in M_{p}\subseteq
725: D_{p}(M,N)=(M\veebar N)_{p}$. Hence $(M\veebar N)(x)\geq p=M(x)$; similarly
726: $(M\veebar N)(x)\geq N(x)$. Since $M\overset{.}{\vee}N=\sup(M,N)$, it follows
727: that $(M\veebar N)(x)\geq(M\overset{.}{\vee}N)(x)$ and so $(M\veebar
728: N)_{p}\supseteq(M\overset{.}{\vee}N)_{p}$.
729:
730: So we have $(M\veebar N)_{p}=(M\overset{.}{\vee}N)_{p}$ which (Proposition
731: \ref{cnt0205}) implies $M\veebar N=M\overset{.}{\vee}N$.\noindent
732: \end{proof}
733:
734: \begin{proposition}
735: \label{cnt0411}$(\widetilde{\mathbf{I}},\leq,\overset{.}{\vee},\wedge)\ $is a
736: distributive lattice.
737: \end{proposition}
738:
739: \begin{proof}
740: We must show that for any $A,B,C\in\widetilde{\mathbf{I}}$ we have
741: $(A\overset{.}{\vee}B)\wedge C$ = $(A\wedge C)\overset{.}{\vee}(B\wedge C)$
742: and $(A\wedge B)\overset{.}{\vee}C$ = $(A\overset{.}{\vee}C)\wedge
743: (B\overset{.}{\vee}C)$. We will show this by showing equality of the $p$-cuts.
744:
745: Indeed, choose any $p\in L$ and set $A_{p}=[a_{1},a_{2}]$, $B_{p}=[b_{1}%
746: ,b_{2}]$, $C_{p}=[c_{1},c_{2}]$ (in case any of these intervals is empty,
747: denote it by $[\sqcap X,\sqcup X]$). Now
748: \[
749: \left( (A\overset{.}{\vee}B)\wedge C\right) _{p}=(A\overset{.}{\vee}%
750: B)_{p}\cap C_{p}=(A_{p}\overset{.}{\cup}B_{p})\cap C_{p}=
751: \]%
752: \[
753: ([a_{1},a_{2}]\overset{.}{\cup}[b_{1},b_{2}])\cap\lbrack c_{1},c_{2}%
754: ]=[a_{1}\sqcap b_{1},a_{2}\sqcup b_{2}]\cap\lbrack c_{1},c_{2}]=
755: \]%
756: \[
757: \lbrack(a_{1}\sqcap b_{1})\sqcup c_{1},(a_{2}\sqcup b_{2})\sqcap
758: c_{2}]=[(a_{1}\sqcup c_{1})\sqcap(b_{1}\sqcup c_{1}),(a_{2}\sqcap c_{2}%
759: )\sqcup(b_{2}\sqcap c_{2}))]=
760: \]%
761: \[
762: \lbrack a_{1}\sqcup c_{1},a_{2}\sqcap c_{2}]\overset{.}{\cup}[b_{1}\sqcup
763: c_{1},b_{2}\sqcap c_{2}]\ =\left( [a_{1},a_{2}]\cap\lbrack c_{1}%
764: ,c_{2}]\right) \overset{.}{\cup}\left( [b_{1},b_{2}]\cap\lbrack c_{1}%
765: ,c_{2}]\right) \ =
766: \]%
767: \[
768: (A_{p}\cap C_{p})\overset{.}{\cup}(B_{p}\cap C_{p})=(A\wedge C)_{p}\overset
769: {.}{\cup}(B\wedge C)_{p}=\left( (A\wedge C)\overset{.}{\vee}(B\wedge
770: C)\right) _{p}.
771: \]
772: Since for all $p\in L$ we have $\left( (A\overset{.}{\vee}B)\wedge C\right)
773: _{p}$ = $\left( (A\wedge C)\overset{.}{\vee}(B\wedge C)\right) _{p}$ , it
774: follows that $(A\overset{.}{\vee}B)\wedge C$ = $(A\wedge C)\overset{.}{\vee
775: }(B\wedge C)$. Dually we show that $(A\wedge B)\overset{.}{\vee}C$ =
776: $(A\overset{.}{\vee}C)\wedge(B\overset{.}{\vee}C)$.
777: \end{proof}
778:
779: \section{Discussion}
780:
781: \label{sec05}
782:
783: In this paper we have introduced fuzzy intervals and obtained some of their
784: basic properties. The method we have used is rather standard in the study of
785: fuzzy algebras -- in particular we have obtained several properties of fuzzy
786: intervals by studying their $p$-cuts. This method can be used to obtain
787: further properties of fuzzy intervals.
788:
789: In our analysis we have made several assumptions, the most prominent ones
790: being that:\ (a)\ $L$ is $[0,1]$ and (b)$\;X$ is complete and completely
791: distributive. To what extent can these assumptions be relaxed?
792:
793: Regarding $L$, the analysis remains unchanged if $(L,\leq,\vee,\wedge)$ is
794: simply a chain. But it does not seem obvious how to generalize our results to
795: L-fuzzy lattices, because Proposition \ref{cnt0407}\ requires that for every
796: $P\subseteq L$, and for all $p,q\in P$, we have $p\vee q\in P$; for this to be
797: true for arbitrary $P\subseteq L$, \ $(L,\leq)$ must be a chain.
798:
799: The completeness of $(X,\sqsubseteq,\sqcup,\sqcap)$ is also essential.
800: Obviously, if $(X,\sqsubseteq,\sqcup,\sqcap)$ is not complete, there is no
801: guarantee that an infinite union of fuzzy intervals will be a fuzzy interval.
802: Regarding complete distributivity, it has only been used in Section
803: \ref{sec04}, but there it plays an essential role in the proof of Proposition
804: \ref{cnt0407}. Let us note that in the important special case where $X$ has
805: finite cardinality, completeness is automatically satisfied and complete
806: distributivity is equivalent to distributivity (which clearly is a minimum
807: requirement for the lattice of fuzzy intervals to be distributive).
808:
809: Finally, let us discuss briefly the important special case when
810: $(X,\sqsubseteq,\sqcup,\sqcap)$ = $(R,\leq,\vee,\wedge)$. In this case we
811: obtain the ``classical'' notion of a fuzzy interval, i.e. a fuzzy set such
812: that its $p$-cuts are closed intervals on the real line (compare \cite[p.37,
813: p.48]{Nguyen}. It is worth noting that, taking $X=R^{n}$, the notion of a
814: fuzzy convex sublattice also specializes to that of a ``classical'' convex
815: fuzzy set \cite[p.41]{Nguyen}. Fuzzy intervals and convex fuzzy sets in this
816: ``classical'' sense have been studied extensively. It appears worthwhile to
817: study ``classical'' fuzzy intervals from the lattice theoretic point of view.
818: Conversely, they can serve as a source of inspiration for generalizations
819: (especially of convexity results) in the context of a general lattice
820: $(X,\sqsubseteq,\sqcup,\sqcap)$.
821:
822: \begin{thebibliography}{9} %
823: \bibitem {AjLat01}N. Ajmal and K.V. Thomas. ``Fuzzy lattices''. \emph{Info.
824: Sciences}. vol. 79, pp.271-291, 1994.
825:
826: \bibitem {AjRng01}N. Ajmal and K.V. Thomas. ``The lattices of fuzzy ideals of
827: a ring''. \emph{Fuzzy Sets and Systems}. vol. 74, pp.371-379, 1995.
828:
829: \bibitem {AjGrp01}N. Ajmal and K.V. Thomas. ``The join of fuzzy algebraic
830: substructures of a group and their lattices''. \emph{Fuzzy Sets and Systems},
831: vol.\emph{\ }99, pp.213--224, 1998.
832:
833: \bibitem {Birkhoff}G. Birkhoff. \emph{Lattice Theory}, American Math. Society,
834: Providence, Rhode Island, 1967.
835:
836: \bibitem {Dixit01}V.N. Dixit, R. Kumar and N. Ajmal. ``On fuzzy rings''.
837: \emph{Fuzzy Sets and Systems}. vol. 49, pp.205-213, 1992.
838:
839: \bibitem {VGK03}V.G. Kaburlasos and V. Petridis. ``Fuzzy Lattice
840: Neurocomputing (FLN) Models''. \emph{Neural Networks}, vol. 13, pp. 1145-1170, 2000.
841:
842: \bibitem {Nguyen}H.T. Nguyen and E.A. Walker. \emph{A First Course on Fuzzy
843: Logic}. CRC\ Press, Boca Raton, 1997.
844:
845: \bibitem {VGK01}V. Petridis V and V. G. Kaburlasos. ``Fuzzy Lattice Neural
846: Network (FLNN): A Hybrid Model for Learning''. \emph{IEEE Transactions on
847: Neural Networks}, vol. 9, pp. 877-890, 1998.
848:
849: \bibitem {VGK02}V. Petridis and V. G. Kaburlasos. ``Learning in the Framework
850: of Fuzzy Lattices''. \emph{IEEE Transactions on Fuzzy Systems.} vol. 7, pp.
851: 422-440, 1999.
852:
853: \bibitem {Ros01}A. Rosenfeld. ``Fuzzy Groups''. \emph{J. Math. Anal. and
854: Appl.,} vol. 35, pp.512-517, 1971.
855:
856: \bibitem {Ses01}B. Seselja and A. Tepavcevic. ``Representation of lattices by
857: fuzzy sets''. \emph{Inform. Sci}., vol. 79, pp.171--180. 1994.
858:
859: \bibitem {Ses03}B. Seselja and A. Tepavcevic. ``On a generalization of fuzzy
860: algebras and congruences''. \emph{Fuzzy Sets and Systems}, vol.65, pp.85--94. 1994.
861:
862: \bibitem {Ses04}B. Seselja. ``Lattice of partially ordered fuzzy
863: subalgebras''. \emph{Fuzzy Sets and Systems}, vol.81, pp.265-269. 1996.
864:
865: \bibitem {Swamy1}U.M. Swamy and D.V. Raju. ``Fuzzy ideals and congruences of
866: lattices''. \emph{Fuzzy Sets and Systems}, vol. 95, pp.249-253, 1998.
867:
868: \bibitem {Tep01}A. Tepavcevic and G. Trajkovski. ``$L$-fuzzy lattices: an
869: introduction''. \emph{Fuzzy Sets and Systems}, vol.\emph{\ }123, pp.209--216. 2001.
870:
871: \bibitem {Wang01}G.Z. Wang. ``Generalized topological molecular lattices''.
872: \emph{Sci. Sinica Ser. A}, vol. 27, pp. 785--798. 1984.
873:
874: \bibitem {Yuan1}B. Yuan and W. Wu. ``Fuzzy ideals on a distributive lattice''.
875: \emph{Fuzzy Sets and Systems}, vol. 35, pp.231-240, 1990.
876:
877: \bibitem {Wang03}L.S. Xu. ``Construct fuzzy lattices from a given symmetric
878: complete lattice''. \emph{Fuzzy Sets and Systems}, vol. 66, pp.357--362. 1994.
879:
880: \bibitem {Yue01}Z. Yue. ``Prime L-fuzzy ideals primary L-fuzzy ideals''.
881: \emph{Fuzzy Sets and Systems}, vol. 27, pp.345-350, 1988.
882:
883: \bibitem {ZahRng01}M.M. Zahedi. ``A characterisation of L-fuzzy prime
884: ideals''. \emph{Fuzzy Sets and Systems}, vol. 44, pp.147-160, 1991..
885:
886: \bibitem {Wang02}X.D. Zhao. ``Representation theorem of fuzzy lattices and its
887: applications in fuzzy topology''. \emph{Fuzzy Sets and Systems},
888: vol.\emph{\ }25, pp.125--128, 1988.
889: \end{thebibliography}
890: \end{document}