Interpretability assessment of fuzzy knowledge bases: A cointension based approach

Computing with words (CWW) relies on linguistic representation of knowledge that is processed by operating at the semantical level defined through fuzzy sets. Linguistic representation of knowledge is a major issue when fuzzy rule based models are acquired from data by some form of empirical learning. Indeed, these models are often requested to exhibit interpretability, which is normally evaluated in terms of structural features, such as rule complexity, properties on fuzzy sets and partitions. In this paper we propose a different approach for evaluating interpretability that is based on the notion of cointension. The interpretability of a fuzzy rule-based model is measured in terms of cointension degree between the explicit semantics, defined by the formal parameter settings of the model, and the implicit semantics conveyed to the reader by the linguistic representation of knowledge. Implicit semantics calls for a representation of user’s knowledge which is difficult to externalise. Nevertheless, we identify a set of properties - which we call “logical view” - that is expected to hold in the implicit semantics and is used in our approach to evaluate the cointension between explicit and implicit semantics. In practice, a new fuzzy rule base is obtained by minimising the fuzzy rule base through logical properties. Semantic comparison is made by evaluating the performances of the two rule bases, which are supposed to be similar when the two semantics are almost equivalent. If this is the case, we deduce that the logical view is applicable to the model, which can be tagged as interpretable from the cointension viewpoint. These ideas are then used to define a strategy for assessing interpretability of fuzzy rule-based classifiers (FRBCs). The strategy has been evaluated on a set of pre-existent FRBCs, acquired by different learning processes from a well-known benchmark dataset. Our analysis highlighted that some of them are not cointensive with user’s knowledge, hence their linguistic representation is not appropriate, even though they can be tagged as interpretable from a structural point of view.     

A construction of sound semantic linguistic scales using 4-tuple representation of term semantics

Data semantics plays a fundamental role in computer science, in general, and in computing with words, in particular. The semantics of words arises as a sophisticated problem, since words being actually vague linguistic terms are pieces of information characterized by impreciseness, incompleteness, uncertainty and/or vagueness. The qualitative semantics and the quantitative semantics are two aspects of vague linguistic information, which are closely related. However, the qualitative semantics of linguistic terms, and even the qualitative semantics of the symbolic approaches, seem to be not elaborated on directly in the literature. In this study, we propose an interpretation of the inherent order-based semantics of terms through their qualitative semantics modeled by hedge algebra structures. The quantitative semantics of terms are developed based on the quantification of hedge algebras. With this explicit approach, we propose two concepts of assessment scales to address decision problems: linguistic scales used for representing expert linguistic assessments and semantic linguistic scales based on 4-tuple linguistic representation model, which forms a formalized structure useful for computing with words. An example of a simple multi-criteria decision problem is examined by running a comparative study. We also analyze the main advantages of the proposed approach.     

An architecture of fuzzy neural networks for linguistic processing

Fuzzy sets have been used successfully in order to deal with imprecise data, linguistic terms or not well-defined concepts. Recently, considerable effort has been made in the direction of combining the neural network approach with fuzzy sets. In this paper a fuzzy feed-forward neural network, able to process trapezoidal fuzzy sets, has been investigated. Normalized trapezoidal fuzzy sets have been considered. The fuzzy generalized delta rule with different back-propagation algorithms is discussed. The more interesting and characteristic property of the proposed architecture is the ability of each node to process fuzzy sets or linguistic terms, preserving the simplicity of the back-propagation algorithm. Consequently, the resulting architecture is able to cope with problems in which the input parameters and the desired targets are described by linguistic terms. This methodology has the further interesting characteristic of being able to operate at the linguistic level rather than at the numerical level, that is it can work at a higher data abstraction level. An example in computerized electrocardiography will be illustrated in order to test the proposed approach.     

Some relationships between fuzzy and random set-based classifiers and models

When designing rule-based models and classifiers, some precision is sacrificed to obtain linguistic interpretability. Understandable models are not expected to outperform black boxes, but usually fuzzy learning algorithms are statistically validated by contrasting them with black-box models. Unless performance of both approaches is equivalent, it is difficult to judge whether the fuzzy one is doing its best, because the precision gap between the best understandable model and the best black-box model is not known.In this paper we discuss how to generate probabilistic rule-based models and classifiers with the same structure as fuzzy rule-based ones. Fuzzy models, in which features are partitioned into linguistic terms, will be compared to probabilistic rule-based models with the same number of terms in every linguistic partition. We propose to use these probabilistic models to estimate a lower precision limit which fuzzy rule learning algorithms should surpass.     

Some mathematical tools for linguistic probabilities

We propose and develop, in this paper, some concepts and techniques useful for the theory of linguistic probabilisies introduced by L.A. Zadeh. These probabilities are expressed in linguistic rather than numerical terms. The mathematical framework for this study is based upon the possibility theory.We formulate first the problem of optimization under elastic constraints which is not only important for mathematical programming but will be served to justify the extension of possibility measure to linguistic variables. Next, in connection with translation rules in natural languages we study some transformations of fuzzy sets using a relation between random sets and fuzzy sets. Finally, we point out some differences between random variables and fuzzy variables, and present the mathematical notion of possibility, in the setting of set-functions, as a special case of Choquet capacities.     

Predicate Modal Logics Do Not Mix Very Well

The problem of completeness for predicate modal logics is still under investigation, although some results have been obtained in the last few years (cf. [2, 3, 4, 7]). As far as we know, the case of multimodal logics has not been addressed at all. In this paper, we study the combination of modal logics in terms of combining their semantics. We demonstrate by a simple example that in this sense predicate modal logics are not so easily manipulated as propositional ones: mixing two Kripke-complete predicate modal logics (one with the Barcan formula, and the other without) results in a Kripke-incomplete system.     

Long-range predictive adaptive fuzzy relational control

A class of long-range predictive adaptive fuzzy relational controllers is presented. The plant behavior is described over an extended time horizon by a fuzzy relational model which is identified based on input-output closed-loop observations of the plant variables. In this class of adaptive controllers the control law attempts to minimize a quadratic cost over an extended control horizon. When used with linear models, this approach has revealed a significant potential for overcoming the limitations of one-step ahead schemes, such as the stabilization of non-minimum phase plants. Here, a uniform framework is adopted for implementing both the fuzzy model and the fuzzy controller, namely distributed fuzzy relational structures gaining from their massive parallel processing features and from the learning capabilities typical of the connectivist approaches. Issues such as maintenance during the adaptation process of the meaning of linguistic terms used at both fuzzy systems interfaces are addressed, namely by introducing a new design methodology for on-line fuzzy systems interface adaptation. The examples presented reinforce the claim of the usefulness of this new approach.     

Improving the accuracy while preserving the interpretability of fuzzy function approximators by means of multi-objective evolutionary algorithms

The identification of a model is one of the key issues in the field of fuzzy system modeling and function approximation theory. An important characteristic that distinguishes fuzzy systems from other techniques in this area is their transparency and interpretability. Especially in the construction of a fuzzy system from a set of given training examples, little attention has been paid to the analysis of the trade-off between complexity and accuracy maintaining the interpretability of the final fuzzy system. In this paper a multi-objective evolutionary approach is proposed to determine a Pareto-optimum set of fuzzy systems with different compromises between their accuracy and complexity. In particular, two fundamental and competing objectives concerning fuzzy system modeling are addressed: fuzzy rule parameter optimization and the identification of system structure (i.e. the number of membership functions and fuzzy rules), taking always in mind the transparency of the obtained system. Another key aspect of the algorithm presented in this work is the use of some new expert evolutionary operators, specifically designed for the problem of fuzzy function approximation, that try to avoid the generation of worse solutions in order to accelerate the convergence of the algorithm.     

Solving maximum fuzzy clique problem with neural networks and its applications

The maximum clique problem is an important problem in graph theory. Many real-life problems are still being mapped into this problem for their effective solutions. A natural extension of this problem that has emerged very recently in many real-life networks, is its fuzzification. The problem of finding the maximum fuzzy clique has been formalized on fuzzy graphs and subsequently addressed in this paper. It has been shown here that the problem reduces to an unconstrained quadratic 0–1 programming problem. Using a maximum neural network, along with mutation capability of genetic adaptive systems, the reduced problem has been solved. Empirical studies have been done by applying the method on stock flow graphs to identify the collusion set, which contains a group of traders performing unfair trading among themselves. Additionally, it has been applied on a gene co-expression network to find out significant gene modules and on some benchmark graphs.     

Applying fuzzy weighted average approach to evaluate office layouts with Feng–Shui consideration

Office layout is an important issue, especially in China and the Asian countries, where the Feng–Shui theory frequently plays a vital role. Yet, in the literature, Feng–Shui theory has seldom been discussed. Another problem is the imprecise or vague satisfaction level of the linguistic expression used in this theory. In this article, the fuzzy set theory is applied to deal with this aspect of the problem. Using an improved and efficient fuzzy weighted average (EFWA) algorithm, which has been shown to be more advantageous than the existing FWA algorithms, an empirical study of an office-layout design problem with the consideration of Feng–Shui is presented to illustrate the EFWA approach. The results and the criteria developed, based on the interpretation of the Form school concept of the Feng–Shui are reported.     

A method based on PSO and granular computing of linguistic information to solve group decision making problems defined in heterogeneous contexts

Group decision making is a type of decision problem in which multiple experts acting collectively, analyze problems, evaluate alternatives, and select a solution from a collection of alternatives. As the natural language is the standard representation of those concepts that humans use for communication, it seems natural that they use words (linguistic terms) instead of numerical values to provide their opinions. However, while linguistic information is readily available, it is not operational and thus it has to be made usable though expressing it in terms of information granules. To do so, Granular Computing, which has emerged as a unified and coherent framework of designing, processing, and interpretation of information granules, can be used. The aim of this paper is to present an information granulation of the linguistic information used in group decision making problems defined in heterogeneous contexts, i.e., where the experts have associated importance degrees reflecting their ability to handle the problem. The granulation of the linguistic terms is formulated as an optimization problem, solved by using the particle swarm optimization, in which a performance index is maximized by a suitable mapping of the linguistic terms on information granules formalized as sets. This performance index is expressed as a weighted aggregation of the individual consistency achieved by each expert.     

Genetic learning of accurate and compact fuzzy rule based systems based on the 2-tuples linguistic representation

One of the problems that focus the research in the linguistic fuzzy modeling area is the trade-off between interpretability and accuracy. To deal with this problem, different approaches can be found in the literature. Recently, a new linguistic rule representation model was presented to perform a genetic lateral tuning of membership functions. It is based on the linguistic 2-tuples representation that allows the lateral displacement of a label considering an unique parameter. This way to work involves a reduction of the search space that eases the derivation of optimal models and therefore, improves the mentioned trade-off.Based on the 2-tuples rule representation, this work proposes a new method to obtain linguistic fuzzy systems by means of an evolutionary learning of the data base a priori (number of labels and lateral displacements) and a simple rule generation method to quickly learn the associated rule base. Since this rule generation method is run from each data base definition generated by the evolutionary algorithm, its selection is an important aspect. In this work, we also propose two new ad hoc data-driven rule generation methods, analyzing the influence of them and other rule generation methods in the proposed learning approach. The developed algorithms will be tested considering two different real-world problems.     

Multidimensional interval-valued fuzzy reasoning approach based on weighted similarity measure

This paper focuses on presentation of a method to bidirectional interval-valued fuzzy approximate reasoning by employing a weighted similarity measure between the fact and the antecedent (or consequent) portion of production rule in which the vague terms are represented by interval-valued fuzzy concepts rather than plain fuzzy sets. The proposed method is more reasonable and flexible than the one presented in the paper by Chen [Fuzzy Sets and Systems, 91(1997), 339–353] due to the fact that it not only can deal with multidimensional interval-valued fuzzy reasoning scheme, but also consider the different importance degree of linguistic variables in production rule and that of elements in each universe.     

Learning concurrently partition granularities and rule bases of Mamdani fuzzy systems in a multi-objective evolutionary framework

In this paper we propose a multi-objective evolutionary algorithm to generate Mamdani fuzzy rule-based systems with different good trade-offs between complexity and accuracy. The main novelty of the algorithm is that both rule base and granularity of the uniform partitions defined on the input and output variables are learned concurrently. To this aim, we introduce the concepts of virtual and concrete rule bases: the former is defined on linguistic variables, all partitioned with a fixed maximum number of fuzzy sets, while the latter takes into account, for each variable, a number of fuzzy sets as determined by the specific partition granularity of that variable. We exploit a chromosome composed of two parts, which codify the variables partition granularities, and the virtual rule base, respectively. Genetic operators manage virtual rule bases, whereas fitness evaluation relies on an appropriate mapping strategy between virtual and concrete rule bases. The algorithm has been tested on two real-world regression problems showing very promising results.     

A historical review of evolutionary learning methods for Mamdani-type fuzzy rule-based systems: Designing interpretable genetic fuzzy systems

The need for trading off interpretability and accuracy is intrinsic to the use of fuzzy systems. The obtaining of accurate but also human-comprehensible fuzzy systems played a key role in Zadeh and Mamdani’s seminal ideas and system identification methodologies. Nevertheless, before the advent of soft computing, accuracy progressively became the main concern of fuzzy model builders, making the resulting fuzzy systems get closer to black-box models such as neural networks. Fortunately, the fuzzy modeling scientific community has come back to its origins by considering design techniques dealing with the interpretability-accuracy tradeoff. In particular, the use of genetic fuzzy systems has been widely extended thanks to their inherent flexibility and their capability to jointly consider different optimization criteria. The current contribution constitutes a review on the most representative genetic fuzzy systems relying on Mamdani-type fuzzy rule-based systems to obtain interpretable linguistic fuzzy models with a good accuracy.     

Bivalent and other solutions of fuzzy relational equations via linguistic hedges

We show that the well-known results regarding solutions of fuzzy relational equations and their systems can easily be generalized to obtain criteria regarding constrained solutions such as solutions which are crisp relations. When the constraint is empty, constrained solutions are ordinary solutions. The generalization is obtained by employing intensifying and relaxing linguistic hedges, conceived in this paper as certain unary functions on the scale of truth degrees. One aim of the paper is to highlight the problem of constrained solutions and to demonstrate that this problem naturally appears when identifying unknown relations. The other is to emphasize the role of linguistic hedges as constraints.     

不确定多属性群决策中混合信息的集结

在不确定多属性群决策中,研究专家给出的评价信息为语言和三角模糊数混合型的决策问题.提出一种转化方法,先将模糊数转化为语言短语集上的模糊集,然后,再将此模糊集转化为二元语义.同时在信息集结过程中,也均以二元语义的形式,以防止信息的过分丢失.最后,给出一个算例来说明此种处理方法的有效性和实用性.     

基于对偶犹豫模糊语言变量的多属性决策方法

首先定义了对偶犹豫模糊语言变量,然后给出其运算规则、得分值函数、精确值函数、比较规则以及对偶犹豫模糊语言变量的加权算术平均算子、有序加权算术平均算子和混合平均算子。针对属性值为对偶犹豫模糊语言变量的多属性决策问题,提出了一种基于对偶犹豫模糊语言变量集结算子的多属性决策方法。最后,结合国家电网公司合作单位选择问题,验证了该方法的有效性和可行性。     

A group decision making model based on goal programming with fuzzy hierarchy: an application to regional forest planning

In this paper a multi-criteria group decision making model is presented in which there is a heterogeneity among the decision makers due to their different expertise and/or their different level of political control. The relative importance of the decision makers in the group is handled in a soft manner using fuzzy relations. We suppose that each decision maker has his/her preferred solution, obtained by applying any of the techniques of distance-based multi-objective programming [compromise, goal programming (GP), goal programming with fuzzy hierarchy, etc.]. These solutions are used as aspiration levels in a group GP model in which the differences between the unwanted deviations are interpreted in terms of the degree of achievement of the relative importance amongst the group members. In this way, a group GP model with fuzzy hierarchy (Group-GPFH) is constructed. The solution for this model is proposed as a collective decision. To show the applicability of our proposal, a regional forest planning problem is addressed. The objective is to determine tree species composition in order to improve the values achieved by Pan-European indicators for sustainable forest management. This problem involves stakeholders with competing interests and different preference schemes for the aforementioned indicators. The application of our proposal to this problem allows us to be able to comfortably address all these issues. The results obtained are consistent with the preferences of each stakeholder and their hierarchy within the group.