Abstract
Our method for density elimination is generalized to the non-commutative substructural logic . Then, the standard completeness of follows as a lemma by virtue of previous work by Metcalfe and Montagna. This result shows that is the logic of pseudo-uninorms and their residua and answered the question posed by Prof. Metcalfe, Olivetti, Gabbay and Tsinakis.
Keywords:
density elimination; pseudo-uninorm logic; standard completeness of HpsUL*; substructural logics; fuzzy logic MSC:
03B50; 03F05; 03B52; 03B47
1. Introduction
Prof. Metcalfe, Olivetti and Gabbay conjectured that the Hilbert system is the logic of pseudo-uninorms and their residua in 2009 in [1]. It is not the case, as shown by Prof. Wang and Zhao in [2], although is the logic of bounded representable residuated lattices. We constructed the system by adding the weakly commutativity rule
to and conjectured that it is the logic of residuated pseudo-uninorms and their residua in 2013 in [3].
In this paper, we prove the conjecture by showing that the density elimination holds for the hypersequent system corresponding to . Then, the standard completeness of follows as a lemma by virtue of previous work by Metcalfe and Montagna [4]. That is, is complete with respect to algebras whose lattice reduct is the real unit interval . Thus, is a kind of substructural fuzzy logic [4], and potentially has certain applications to fuzzy inferences and expert Systems [5,6,7,8]. Our result also shows that that is an axiomatization for the variety of residuated lattices generated by all dense residuated chains. Thus, we have also answered the question posed by Prof. Metcalfe and Tsinakis in [9] in 2017.
In proving the density elimination for , we have to overcome several difficulties as follows. Firstly, cut-elimination doesn’t holds for . Note that and the density rule are formulated as
in , respectively. Consider the following derivation fragment.
By the induction hypothesis of the proof of cut-elimination, we get that
from and by . However, we can’t deduce from by . We overcome this difficulty by introducing the following weakly cut rule into
Secondly, the proof of the density elimination for becomes troublesome even for some simple cases in [4]. Consider the following derivation fragment
Here, the major problem is how to extend such that it is applicable to . By replacing p with t, we get . However, there exists no derivation of from and . Notice that and in are commutated simultaneously in , which we can’t obtain by . It seems that can’t be strengthened further in order to solve this difficulty. We overcome this difficulty by introducing a restricted subsystem of . is a generalization of , which we introduced in [10] in order to solve a longstanding open problem, i.e., the standard completeness of . Two new manipulations, which we call the derivation-splitting operation and derivation-splicing operation, are introduced in .
The third difficulty we encounter is that the conditions of applying the restricted external contraction rule become more complex in because new derivation-splitting operations make the conclusion of the generalized density rule to be a set of hypersequents rather than one hypersequent. We continue to apply derivation-grafting operations in the separation algorithm of the multiple branches of in [10], but we have to introduce a new construction method for by induction on the height of the complete set of maximal -nodes rather than on the number of branches.
The structure of this paper is as follows. In Section 2, we present two hypersequent calculi and , and prove that Cut-elimination does not hold for . Because of the absence of the commutativity rule, we have to introduce two novel operations, i.e., the derivation-splitting operation and derivation-splicing operation, in in Section 3, and then we present a suitable definition of the generalized density rule for . In Section 4, we adapt the old main algorithm in the system to the new system . In Section 5, we propose two directions for future research.
2. , and
Definition 1.
([1]) consists of the following initial sequents and rules:
Initial sequents
Structural Rules
Logical Rules
Cut Rule
Definition 2.
([3]) is plus the weakly commutativity rule
Definition 3.
is plus the density rule .
Lemma 1.
is not a theorem in .
Proof.
Let be an algebra, where , for all , and the binary operations ⊙, → and ⇝ are defined by the following tables (see [2]).
By easy calculation, we get that is a linearly ordered -algebra, where 0 and 3 are the least and the greatest element of , respectively, and 2 is its unit. Let . Then, . Hence, G is not a tautology in . Therefore, it is not a theorem in by Theorem 9.27 in [1]. □
Theorem 1.
Cut-elimination doesn’t hold for .
Proof.
is provable in , as shown in Figure 1.
Figure 1.
A proof of G.
Suppose that G has a cut-free proof . Then, there exists no occurrence of t in by its subformula property. Thus, there exists no application of in . Hence, G is a theorem of , which contradicts Lemma 1. □
Remark 1.
Following the construction given in the proof of Theorem 53 in [4], in Figure 1 is eliminated by the following derivation, as shown in Figure 2. However, the application of in ρ is invalid, which illustrates the reason why the cut-elimination theorem doesn’t hold in .
Figure 2.
A possible cut-free proof ρ of G.
Definition 4.
is constructed by replacing in with
We call it the weakly cut rule and denote it by .
Theorem 2.
If , then .
Proof.
It is proved by a procedure similar to that of Theorem 53 in [4] and omitted. □
Definition 5.
( [10]) is a restricted subsystem of such that
(i) p is designated as the unique eigenvariable by which we mean that it is not used to build up any formula containing logical connectives and is only used as a sequent-formula.
(ii) Each occurrence of p in a hypersequent is assigned one unique identification number i in and written as . Initial sequent of has the form in . p doesn’t occur in or Δ for each initial sequent or in .
(iii) Each sequent S of the form in has the form in , where p does not occur in for all and, for all . Define , if A is an eigenvariable with the identification number and, if A isn’t an eigenvariable.
Let G be a hypersequent of in the form then and for all . Define , .
(iv) A hypersequent G of is called closed if . Two hypersequents and of are called disjoint if , , and . is a copy of if they are disjoint and there exist two bijections and such that can be obtained by applying to antecedents of sequents in and to succedents of sequents in .
(v) A hypersequent can be contracted as in under certain conditions given in Construction 3, which we called the constraint external contraction rule and denote by .
(vi) is forbidden in and and are replaced with and , respectively.
(vii) Two rules and of are replaced with and in , respectively.
(viii) and are closed and disjoint for each two-premise inference rule of and, is closed for each one-premise inference rule .
Proposition 1.
Let and be inference rules of Then, and .
Proof.
Although makes t’s in its premises disappear in its conclusion; it has no effect on identification numbers of the eigenvariable p in a hypersequent because t is a constant in and is distinguished from propositional variables. □
Definition 6.
Let G be a closed hypersequent of and . is called a minimal closed unit of G.
3. The Generalized Density Rule for
In this section, is without . Generally, , denote a formula other than an eigenvariable .
Construction 1.Given a proof of in , let , where , . By , we denote the sequent containing in . Then, , , and . Hypersequents , and their proofs , are constructed inductively for all in the following such that , , and .
(i) , and are built up with .
(ii) Let (or be in , and (accordingly for for some . There are three cases to be considered.
Case 1.
. If all focus formula(s) of is (are) contained in ,
(accordingly for ) and, is constructed by combining the derivation and (accordingly for and, is constructed by combining and . The case of all focus formula(s) of contained in is dealt with by a procedure dual to above and omitted.
Case 2.
. (accordingly for , and is constructed by combining the derivation and (accordingly for and, is constructed by combining and .
Case 3.
. It is dealt with by a procedure dual to Case 2 and omitted.
Definition 7.
The manipulation described in Construction 1 is called the derivation-splitting operation when it is applied to a derivation and the splitting operation when applied to a hypersequent.
Corollary 1.
Let . Then, there exist two hypersequents and such that , , and .
Construction 2.
Given a proof of in , let , where and . Then, there exists such that is in the form and is in the form . A proof of in is constructed by induction on as follows:
- For the base step, let . Then, , where and and and . It follows from Corollary 1 that there exist and such that , , and . Then, is proved as follows:
- For the induction step, let . Then, it is treated using applications of the induction hypothesis to the premise followed by an application of the relevant rule. For example, let , where and and . By the induction hypothesis, we obtain a derivation of :
Definition 8.
The manipulation described in Construction 2 is called the derivation-splicing operation when it is applied to a derivation and the splicing operation when applied to a hypersequent.
Corollary 2.
If , then .
Definition 9.
(i) Let . Define , and , where and are determined by Corollary 1.
(ii) Let . Define .
(iii) Let . if does not occur in G.
(iv) Let for all . Define .
(v) Let and . Define . Especially, define .
Theorem 3.
Let . Then, for all .
Proof.
Immediately from Corollaries 1, 2 and Definition 9. □
Lemma 2.
Let be a minimal closed unit of . Then, has the form if there exists one sequent such that A is not an eigenvariable otherwise has the form .
Proof.
Define in Construction 5.2 in [10]. Then, . Suppose that is constructed such that . If , the procedure terminates and ; otherwise, and define to be an identification number in . Then, there exists by and define Thus, . Hence, there exists a sequence of identification numberssuch that for all , where , for all . Therefore, has the form . □
Definition 10.
Let be a minimal closed unit of . is a splicing unit if it has the form . is a splitting unit if it has the form .
Lemma 3.
Let be a splicing unit of in the form and . Then, .
Proof.
By the construction in the proof of Lemma 2, for all . Then, and , where is obtained by replacing in with . Then, . Repeatedly, we get . □
This shows that is constructed by repeatedly applying splicing operations.
Definition 11.
Let be a minimal closed unit of . Define , and, j is called the child node of i for all . We call the Ω-graph of .
Let be a splitting unit of in the form . Then, each node of has one and only one child node. Thus, there exists one cycle in by . Assume that, without loss of generality, is the cycle of . Then, , , and . Thus, is in the form . By a suitable permutation of , we get . This process also shows that there exists only one cycle in . Then, we introduce the following definition.
Definition 12.
(i) is called a splitting sequent of and its corresponding splitting variable for all .
(ii) Let and . Define , and .
Lemma 4.
If be a splitting unit of , and k be a splitting variable of . Then, is constructed by repeatedly applying splicing operations and only the last operation is a splitting operation.
Construction 3
(The constrained external contraction rule).Let , and be two copies of a minimal closed unit , where we put two copies into and in order to distinguish them. For any splitting unit , or , where . Then, is constructed by cutting off and some sequents in as follows.
(i) If and are two splicing units, then ;
(ii) If and are two splitting units and, k, their splitting variables, respectively, , , , , or , where and is a copy of . Then, .
The above operation is called the constrained external contraction rule, denoted by and written as .
Lemma 5.
If as above, then for all .
4. Density Elimination for
In this section, we adapt the separation algorithm of branches in [10] to and prove the following theorem.
Theorem 4.
Density elimination holds for .
The proof of Theorem 4 runs as follows. It is sufficient to prove that the following strong density rule
is admissible in , where , p does not occur in for all , .
Let be a proof of in by Theorem 2. Starting with , we construct a proof of in by a preprocessing of described in Section 4 in [10].
In Step 1 of preprocessing of , a proof is constructed by replacing inductively all applications of and in with and followed by an application of , respectively. In Step 2, a proof is constructed by converting all into , where . In Step 3, a proof is constructed by converting into , where . In Step 4, a proof is constructed by replacing some (or ) with (or . In Step 5, a proof is constructed by assigning the unique identification number to each occurrence of p in . Let denote the unique node of such that and is the focus sequent of in . We call , the i-th -node of and -sequent, respectively. If we ignore the replacements from Step 4, each sequent of G is a copy of some sequent of and each sequent of is a copy of some contraction sequent in .
Now, starting with and its proof , we construct a proof of in such that each sequent of is a copy of some sequent of G. Then, by Theorem 3 and Lemma 5. Then, by Lemma 9.1 in [10].
In [10], is constructed by eliminating -sequents in one by one. In order to control the process, we introduce the set of maximal -nodes of (see Definition 13) and the set of the branches relative to I and construct such that doesn’t contain the contraction sequents lower than any node in I, i.e., implies for all . The procedure is called the separation algorithm of branches in [10].
The problem we encounter in is that Lemma 7.11 of [10] doesn’t hold because new derivation-splitting operations make the conclusion of -rule to be a set of hypersequents rather than one hypersequent. Then, generally can’t be contracted to in Step 2 of Stage 1 in the main algorithm in [10] and can’t be contracted to in Step 2 of Stage 2. Furthermore, we sometimes can’t construct some branches to I in before we construct . Therefore, we have to introduce a new induction strategy for and don’t perform the induction on the number of branches. First, we give some primary definitions and lemmas.
Definition 13.
A -node is maximal if no other -node is higher than . Define to be the set of maximal -nodes in . A nonempty subset I of is complete if I contains all maximal -nodes higher than or equal to the intersection node of I. Define if , i.e., the intersection node of a single node is itself.
Proposition 2.
(i) for all , .
(ii) Let I be complete and . Then, for some .
(iii) is complete and is complete for all .
(iv) If is complete and , then and are complete, where and denote the sets of all maximal -nodes in the left subtree and right subtree of , respectively.
(v) If are complete, then , or .
Proof.
Only (v) is proved as follows. , or holds by , or , respectively. □
Definition 14.
A labeled binary tree ρ is constructed inductively by the following operations:
(i) The root of ρ is labeled by and leaves labeled .
(ii) If an inner node is labeled by I, then its parent nodes are labeled by and , where and are defined in Proposition 2(iv).
Definition 15.
We define the height of by letting for each leave and, for any non-leaf node.
Note that in Lemma 7.11 in [10] only uniqueness of in doesn’t hold in and the following lemma holds in .
Lemma 6.
Let , , . Then, is separable in and there are some copies of in .
Lemma 7.
(New main algorithm for GpsULΩ)Let I be a complete subset of and . Then, there exists one close hypersequent and its derivation in such that
(i) is constructed by initial hypersequent , the fully constraint contraction rules of the form and elimination rule of the form
where for all , , , , is closed for all . Then, for each and .
(ii) For all , let
where is the skeleton of , which is defined by Definition 7.13 [10]. Then, for some in ;
(iii) Letting and , then and it is built up by applying the separation algorithm along to H, and is an upper hypersequent of either if it is applicable, or , otherwise.
(iv) implies for all and, for some .
Proof.
is constructed by induction on . For the base case, let ; then, is built up by Construction 7.3 and 7.7 in [10]. For the induction case, suppose that , and are constructed such that Claims from (i) to (iv) hold.
Let , where . Then, and occur in the left subtree and right subtree of , respectively. Here, almost all manipulations of the new main algorithm are the same as those of the old main algorithm. There are some caveats that need to be considered.
Firstly, all leaves are replaced with in Step 3 at Stage 1 in the old main algorithm and are replaced with in Step 3 at Stage 2. Secondly, we abandon the definitions of branch to I and Notation 8.1 in [10] and then the symbol of the set of branches, which occur in in [10], is replaced with I in the new algorithm. We call the new algorithm the separation algorithm along I. We also replace in with ☆. Thirdly, under the new requirement that I is complete, we prove the following property.
Property (A) contains at most one copy of .
Proof. Suppose that there exist two copies and of in , and we put them into and in order to distinguish them. Let be a splitting unit of and S its splitting sequent. Then, . Thus, S is a -sequent and has the form by . Then, , for all and for some by Claim (iv). Since is complete and , then .
Let be in the form where , , is the intersection node of and , as shown in Figure 3. Then, by and . Since is separable in by , then and is not .
Figure 3.
A fragment of .
Before proceeding to prove Property (A), we present the following property of .
Property (B) The set of splitting sequents of is equal to that of .
Proof.
Let , and . Then, and are separable in . Thus, is closed. Hence, is closed, where in runs over all above such that . Therefore, , and . Then, the set of splitting sequents of is equal to that of since each splitting sequent is a -sequent by and . This completes the proof of Property (B). □
We therefore assume that, without loss of generality, is in the form by Property (B), Lemma 5 and the observation that each derivation-splicing operation is local. There are two cases to be considered in the following.
Case 1.
for all , . Then, . We assume that, without loss of generality, , . Then, since isn’t a focus sequent at all nodes from to in and, or for all by Lemma 6.7 in [10]. Thus, . Therefore, because , and . This shows that any splitting unit outside in doesn’t take two copies of apart, i.e., the case of and doesn’t happen.
Case 2.
for some , . Then, . Thus, . Hence, . The case of is tackled with the same procedure as the following. Let . Then, there exists a copy of in and let be its splitting sequent. We put two splitting units into and in order to distinguish them. Then, and . We assume that, without loss of generality, , . Then, . Thus, by . Then, , , where we put two copies of into and in order to distinguish them. Then, , , and is a copy of . Then, could be cut off of one of them because they are the two same sets of hypersequents in . Meanwhile, two copies of in can’t be taken apart by any splitting unit outside in for the reason as shown in Case 1 and thus could be contracted into one by in . Therefore, two copies and of can be contracted into one in by . This completes the proof of Property (A). □
With Property (A), all manipulations in the old main algorithm in [10] work well. This completes the construction of and the proof of Theorem 4. □
Theorem 5.
The standard completeness holds for .
Proof.
Let denote the i-th logical link of iff in the following. means that for every algebra in and valuation v on . Let , , and denote the classes of all -algebras, -chain, dense -chain and standard -algebras (i.e., their lattice reducts are ), respectively. We have an inference sequence, as shown in Figure 4.
Figure 4.
Two ways to prove standard completeness.
Links from 1 to 4 show Jenei and Montagna’s algebraic method to prove standard completeness and, currently, it seems hopeless to build up link 3 (see [11,12,13,14]). Links from to show Metcalfe and Montagna’s proof-theoretical method. Density elimination is at Link in Figure 4 and other links are proved by standard procedures with minor revisions and omitted (see [1,4,15,16,17]). □
5. Future Works
Generally, for any existing fuzzy logic system, we can consider its corresponding non-commutative system, just as is obtained by removing the commutativity of the strong conjunctive connective ⊙ in . Therefore, we can consider the corresponding non-commutative systems of many systems. A natural question is whether the method of the density elimination proposed in this paper can be generalized to these systems. It has often been the case in the past that metamathematical methods have corresponding algebraic analogues. The method proposed in this paper is essentially proof-theoretic. A natural problem is whether there is an algebraic proof corresponding to our proof-theoretic one.
Funding
This research was funded by the National Foundation of Natural Sciences of China (Grant No: 61379018, 61662044, 11571013, and 11671358).
Conflicts of Interest
The authors declare no conflict of interest.
References
- Metcalfe, G.; Olivetti, N.; Gabbay, D. Proof Theory for Fuzzy Logics; Springer Series in Applied Logic; Springer: Berlin, Germany, 2009; Volume 36, ISBN 9781402094095. [Google Scholar]
- Wang, S.M.; Zhao, B. HpsUL is not the logic of pseudo-uninorms and their residua. Log. J. IGPL 2009, 17, 413–419. [Google Scholar] [CrossRef] [Scilit]
- Wang, S.M. Logics for residuated pseudo-uninorms and their residua. Fuzzy Sets Syst. 2013, 218, 24–31. [Google Scholar] [CrossRef] [Scilit]
- Metcalfe, G.; Montagna, F. Substructural fuzzy logics. J. Symb. Log. 2007, 7, 834–864. [Google Scholar] [CrossRef] [Scilit]
- Moallem, P.; Mousavi, B.S.; Naghibzadeh, S.S. Fuzzy inference system optimized by genetic algorithm for robust face and pose detection. Int. J. Artif. Intell. 2015, 13, 73–88. [Google Scholar]
- Jankowski, J.; Kazienko, P.; Watróbski, J.; Lewandowska, A.; Ziemba, P.; Zioło, M. Fuzzy multi-objective modeling of effectiveness and user experience in online advertising. Expert Syst. Appl. 2016, 65, 315–331. [Google Scholar] [CrossRef] [Scilit]
- Precup, R.E.; Tomescu, M.L.; Preitl, Ş. Fuzzy logic control system stability analysis based on Lyapunov’s direct method. Int. J. Comput. Commun. Control 2009, 4, 415–426. [Google Scholar] [CrossRef] [Scilit]
- Škrjanc, I.; Blažič, S.; Matko, D. Direct fuzzy model-reference adaptive control. Int. J. Intell. Syst. 2002, 17, 943–963. [Google Scholar] [CrossRef] [Scilit]
- Metcalfe, G.; Tsinakis, C. Density revisited. Soft Comput. 2017, 21, 175–189. [Google Scholar] [CrossRef] [Scilit]
- Wang, S.M. Density Elimination for Semilinear Substructural Logics. arXiv, 2015; arXiv:1509.03472. [CrossRef] [Scilit]
- Jenei, S.; Montagna, F. A proof of standard completeness for Esteva and Godo’s logic MTL. Stud. Log. 2002, 70, 183–192. [Google Scholar] [CrossRef] [Scilit]
- Wang, S.M. Uninorm logic with the n-potency axiom. Fuzzy Sets Syst. 2012, 205, 116–126. [Google Scholar] [CrossRef] [Scilit]
- Wang, S.M. Involutive uninorm logic with the n-potency axiom. Fuzzy Sets Syst. 2013, 218. [Google Scholar] [CrossRef] [Scilit]
- Wang, S.M. The Finite Model Property for Semilinear Substructural Logics. Math. Log. Q. 2013, 59, 268–273. [Google Scholar] [CrossRef] [Scilit]
- Ciabattoni, A.; Galatos, N.; Terui, K. Algebraic proof theory for substructural logics: Cut-elimination and completions. Ann. Pure Appl. Log. 2012, 163, 266–290. [Google Scholar] [CrossRef] [Scilit]
- Nikolaos, G.; Jipsen, P.; Kowalski, T.; Ono, H. Residuated Lattices: An Algebraic Glimpse at Substructural Logics; Elsevier: Amsterdam, The Netherlands, 2007; ISBN 978-0-444-52141-5. [Google Scholar]
- Baldi, P.; Terui, K. Densification of FL chains via residuated frames. Algebr. Univers. 2016, 75, 169–195. [Google Scholar] [CrossRef] [Scilit]
© 2019 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).



