Information Processing and Management of Uncertainty in Knowledge-Based Systems

The interest of the decision makers in the selection process of suppliers is constantly growing as a reliable supplier reduces costs and improves the quality of products/services. This process is essentially reducible to the problem of multi-attribute decision-making. Namely, the large number of quantitative and qualitative attributes is considered. This paper presents a model of supplier selection. Weighted approach for solving this model was used combined with logical interactions between attributes. Setting logical conditions between attributes was carried out by using the Boolean Interpolative Algebra. Then the logical conditions are transformed into generalized Boolean polynomial that is through logical aggregation translated into a single value. In this way, the ranking of the suppliers is provided. Using this model managers will be able to clearly express their demands through logical conditions, i.e. will be able to conduct a comprehensive analysis of the problem and to make an informed decision.

We describe one probabilistic approach to classification of a set of objects when a classification criterion can be represented as a propositional formula. It is well known that probability measures are not truth functional. However, if

μ

is any probability measure and

α

is any propositional formula,

μ

(

α

) is uniquely determined by the

μ

-values of conjunctions of pairwise distinct propositional letters appearing in

α

. In order to infuse truth functionality in the generation of finitely additive probability measures, we need to find adequate binary operations on [0,1] that will be truth functions for finite conjunctions of pairwise distinct propositional letters. The natural candidates for such truth functions are t-norms. However, not all t-norms will generate a finitely additive probability measure. We show that Gödel’s t-norm and product t-norm, as well as their linear convex combinations, can be used for the extension of any evaluation of propositional letters to finitely additive probability measure on formulas. We also present a software for classification of patients with suspected systemic erythematosus lupus (SLE), which implements the proposed probabilistic approach.

This paper demonstrates the application possibilities of the generalised model for Logical Aggregation (LA) in performance and quality evaluation of Green buildings through establishing a linear order of them by a (scalar) general aggregated quality parameter and sorting them into categories. An existing classical fuzzy model is experimentally modified in order to demonstrate the advanced capabilities of adequate articulation of the partial demands for quality via the new generalised model for logical aggregation.

The value of the complex Boolean function can be calculated directly on the basis of its components value. It is a principle known as the truth functionality. Properties of the Boolean algebra have indifferent values. The truth functional principle is taken as a valid principle in general case in the conventional generalization: multi-valued and/or real-valued realizations (fuzzy logic in the broad sense). This paper presents that truth functionality is not valued indifferent property of the Boolean algebra and it is valid only in two-valued realization, and thus it cannot be the basic of the value generalization. The value generalization (real-valued realizations) enables incomparably more descriptiveness than the two-valued classical Boolean algebra, so that the finite Boolean algebra is enough for any real application. Each finite Boolean algebra is atomic. Every Boolean function (the element of the analyzed finite Boolean algebra) can be presented uniquely as disjunction of the relevant atoms – disjunctive canonical form. Which atoms are and which are not included in the analyzed Boolean function is defined by its structure: 0-1 vector which dimension matches the number of atoms (in the case of n independent variables, the number of atoms is 2

n

). Atom corresponds uniquely to each vector structure position and value 0 means that the adequate atom is not included in the analyzed function, and 1 means that it is included. The principle of the structural functionality is: the structure of the complex Boolean function is defined directly on the basis of its components structure. The truth functionality is a value image of the structural functionality only in the case of two-valued realization. Each insisting on the truth functionality, such as in the case of conventional multi-valued logic and fuzzy logic in general sense, is unjustified from the point of the Boolean consistency.

In the study of the Social Networks, the Small World phenomenon appears frequently. We apply some techniques of graph theory and fuzzy sets to characterize the Small World features as well as the existence of the figure of leader in Social Networks. These techniques help to the conceptual formalization in relational networks analysis, by transforming linguistic and human-focused manner concepts related to social networks in some formal representation. These techniques are also applied when the similarity among nodes wants to be measured in order to study the current homophily present in a Network.

Social Network Analysis in the last decade has gained remarkable attention. The current analysis focuses more and more on the dynamic behavior of them. The underlying structure from Social Networks, like facebook, or twitter, can change over time. Groups can be merged or single nodes can move from one group to another. But these phenomenas do not only occur in social networks but also in human brains. The research in neural spike trains also focuses on finding functional communities. These communities can change over time by switching the stimuli presented to the subject. In this paper we introduce a data generator to create such dynamic behavior, with effects in the interactions between nodes. We generate time stamps for events for one-to-one, one-to-many, and many-to-all relations. This data could be used to demonstrate the functionality of algorithms on such data, e.g. clustering or visualization algorithms. We demonstrated that the generated data fulfills common properties of social networks.

The problem of preserving privacy in recommendation systems is faced in this work. The approach presented reduces the study of privacy threats to the study of frequent property set obtained from the characteristics of the objects the recommendation system provides to a target user. This study is made by defining a prominence index for each item and by using efficient methods to explore the lattice of item characteristics.

Nowadays, social networks such as Twitter, Facebook and LinkedIn become increasingly popular. In fact, they introduced new habits, new ways of communication and they collect every day several information that have different sources. Most existing research works focus on the analysis of homogeneous social networks,

i.e.

we have a single type of node and link in the network. However, in the real world, social networks offer several types of nodes and links. Hence, with a view to preserve as much information as possible, it is important to consider social networks as heterogeneous and uncertain. The goal of our paper is to classify the social message based on its spreading in the network and the theory of belief functions. The proposed classifier interprets the spread of messages on the network, crossed paths and types of links. We tested our classifier on a real word network that we collected from Twitter, and our experiments show the performance of our belief classifier.

Assessing the similarity between node profiles in a social network is an important tool in its analysis. Several approaches exist to study profile similarity, including semantic approaches and natural language processing. However, to date there is no research combining these aspects into a unified measure of profile similarity. Traditionally, semantic similarity is assessed using keywords, that is, formatted text information, with no natural language processing component. This study proposes an alternative approach, whereby the similarity assessment based on keywords is applied to the output of natural language processing of profiles. A

unified similarity measure

results from this approach. The approach is illustrated on a real data set extracted from Facebook and compared with other similarity measures for the same data.

In its perpetual capacity to imagine, create and revisit artifacts and representations, human mind is the perfect example of generativity. Yet if we agree with Epstein (1996)’s theory of generativity, new ideas result from interconnections among old ones. That is, cultural knowledge heavily influences our individual minds. In this line, our minds need meeting other minds to generate innovation. I will argue in this article that the basis of a generative meeting between minds is imitation. This proposal is developed against the well-established reputation of imitation as an idiotic behaviour stifling creativity.

We present two axiomatic and three conjectural conditions which a model inducing natural language categories should dispose of, if ever it aims to be considered as “cognitively plausible”. 1st axiomatic condition is that the model should involve a bootstrapping component. 2nd axiomatic condition is that it should be data-driven. 1st conjectural condition demands that the model integrates the surface features – related to prosody, phonology and morphology – somewhat more intensively than is the case in existing Markov-inspired models. 2nd conjectural condition demands that asides integrating symbolic and connectionist aspects, the model under question should exploit the global geometric and topologic properties of vector-spaces upon which it operates. At last we shall argue that model should facilitate qualitative evaluation, for example in form of a POS-i oriented Turing Test. In order to support our claims, we shall present a POS-induction model based on trivial k-way clustering of vectors representing suffixal and co-occurrence information present in parts of Multext-East corpus. Even in very initial stages of its development, the model succeeds to outperform some more complex probabilistic POS-induction models for lesser computational cost.

This paper presents an approach for action recognition performed by human using the joint angles from skeleton information. Unlike classical approaches that focus on the body silhouette, our approach uses body joint angles estimated directly from time-series skeleton sequences captured by depth sensor. In this context, 3D joint locations of skeletal data are initially processed. Furthermore, the 3D locations computed from the sequences of actions are described as the angles features. In order to generate prototypes of actions poses, joint features are quantized into posture visual words. The temporal transitions of the visual words are encoded as symbols for a Hidden Markov Model (HMM). Each action is trained through the HMM using the visual words symbols, following, all the trained HMM are used for action recognition.

The main question addressed in this work deals with the difficulty to compare different data trajectories patterns in particular due to the non symmetry properties. In order to tackle this fundamental issue in its generality from a theoretical point of view, we introduce the quasipseudo-metrics concepts, and with the induced pre-topological space on the datasets we can identify proximity between trajectories. We will illustrate the ideas by discussing the application of the theoretical framework on gestual analysis.

Museums have large databases of images. The librarians that are using these databases are doing two types of images search: either they know what they are looking for in the database (a specific image or a specific set of well defined images such as kings of France), or they do not know precisely what they are looking for (e.g., when they are required to build images portfolios about some concepts such as “decency”). As each image is having a number of metadata, searching for a well-defined image, or for set of images, is easily solved. On the contrary, this is a hard problem when the task is to illustrate a given concept such as “freedom”, “decency”, “bread”, or “transparency” since these concepts are not metadata. How to find images that are somewhat analogs because they illustrate a given concept?

We collected and analyzed the search results of librarians that were given themselves the task of finding images related to a given concept. Seven relations between the concept and the images were found as explanation of the selection of images for any concept: conceptual property, causality, effectivity semantic, anti-logic, metaphorical-vehicle and metaphorical-topic. The inter-rate agreement of independent judges that evaluated the relations was of .78.

Finally, we designed an experiment to evaluate how much metaphor in images can be understandable.

In this paper, the application of the fuzzy transforms of the zero degree (

F

0

-transform) and of the first degree (

F

1

-transform) to the image registration is demonstrated. The main idea is to use only one technique (F-transform generally) to solve various tasks of the image registration. The

F

1

-transform is used for an extraction of feature points in edge detection step. The correspondence between the feature points in two images is obtained by the image similarity algorithm based on the

F

0

-transform. Then, the shift vector for corresponding corners is computed, and by the image fusion algorithm, the final image is created.

The article summarizes current approaches to image fusion problem using fuzzy transformation (F-transform) with their weak points and proposes improved version of the algorithm which suppress them. The first part of this contribution brings brief theoretical introduction into problem domain. Next part analyses weak points of current implementations. Last part introduces improved algorithm and compares it with the previous ones.

Images convey multiple meanings that depend on the context in which the viewer perceptually organizes the scene. By assuming a standardized natural-scene-perception-taxonomy comprised of a hierarchy of nested spatial-taxons [17] [6] [5], image segmentation is operationalized into a series of two-class inferences. Each inference determines the optimal spatial-taxon region, partitioning a scene into a foreground, subject and salient objects and/or sub-objects. I demonstrate the results of a fuzzy-logic-natural-vision-processing engine that implements this novel approach. The engine uses fuzzy-logic inference to simulate low-level visual processes and a few rules of figure-ground perceptual organization. Allowed spatial-taxons must conform to a set of ”meaningfulness” cues, as specified by a generic scene-type. The engine was tested on 70 real images composed of three ”generic scene-types”, each of which required a different combination of the perceptual organization rules built into our model. Five human subjects rated image-segmentation quality on a scale from 1 to 5 (5 being the best). The majority of generic-scene-type image segmentations received a score of 4 or 5 (very good, perfect). ROC plots show that this engine performs better than normalized-cut [9] on generic-scene type images.

This paper proposes two complementary color systems:

red-green-blue-white-black

and

cyan-magenta-yellow-black-white

. Both systems belong to the five-valued category and they represent some particular case of neutrosophic information representation. The proposed multi-valued fuzzy spaces are obtained by constructing fuzzy partitions in the unit cube. In the structure of these five-valued representations, the negation, the union and the intersection operators were defined. Next, using the proposed multi-valued representation in the framework of fuzzy clustering algorithm, it results some color image clustering procedure.

A new fuzzy edge detector based on uninorms is proposed and deeply studied. The behaviour of different classes of uninorms is discussed. The obtained results suggest that the best uninorm in order to improve the edge detection process is the uninorm

$\mathcal{U}_{\rm{min}}$

, with underlying Łukasiewicz operators. This algorithm gets statistically substantial better results than the others obtained by well known edge detectors, as Sobel, Roberts and Prewitt approaches and comparable to the results obtained by Canny.

Many approaches for consistency restoration in database systems have to deal with the problem of an exponential blowup in the number of possible repairs. For this reason, recent approaches advocate more flexible and fine grained policies based on the reasoner’s preference. In this paper we take a further step towards more personalized inconsistency management by incorporating ideas from context-aware systems. The outcome is a general distance-based approach to inconsistency maintenance in database systems, controlled by context-aware considerations.

In the aim of evaluating and improving link quality in bibliographical knowledge bases, we develop a decision support system based on partitioning semantics. The novelty of our approach consists in using symbolic values criteria for partitioning and suitable partitioning semantics. In this paper we evaluate and compare the above mentioned semantics on a real qualitative sample. This sample is issued from the catalogue of French university libraries (SUDOC), a bibliographical knowledge base maintained by the University Bibliographic Agency (ABES).

Data exchange is concerned with the transformation of data structured under one schema into a different schema. In practice, this task is usually accomplished in a procedural way. In a landmark paper, Fagin et al. have proposed a declarative, purely logical approach to this task. Since then, data exchange has been intensively studied in the database research community. Recently, it has been extended to probabilistic data exchange. In this paper, we propose an extension to fuzzy data exchange.

Fuzzy relational compositions have been extensively studied by many authors. Especially, we would like to highlight initial studies of the fuzzy relational compositions motivated by their applications to medical diagnosis by Willis Bandler and Ladislav Kohout. We revisit these types of compositions and introduce new definitions that directly employ generalized quantifiers. The motivation for this step is twofold: first, the application needs for filling a huge gap between the classical existential and universal quantifiers and second, the already existing successful implementation of generalized quantifiers in so called divisions of fuzzy relations, that constitute a database application counterpart of the theory of fuzzy relational compositions. Recall that the latter topic is studied within fuzzy relational databases and flexible querying systems for more than twenty years. This paper is an introductory study that should demonstrate a unifying theoretical framework and introduce that the properties typically valid for fuzzy relational compositions are valid also for the generalized ones, yet sometimes in a weaken form.

In this contribution, we present a functional approach to the cardinality of finite fuzzy sets, it means an approach based on one-to-one correspondences between fuzzy sets. In contrast to one fixed universe of discourse used for all fuzzy sets, our theory is developed within a class of fuzzy sets which universes of discourse are countable sets, and finite fuzzy sets are introduced as fuzzy sets with finite supports. We propose some basic operations with fuzzy sets as well as two constructions - fuzzy power set and fuzzy exponentiation. To express the fact that two finite fuzzy sets have approximately the same cardinality we propose the concept of graded equipollence. Using this concept we provide graded versions of several well-known statements, including the Cantor-Bernstein theorem and the Cantor theorem.

A reasonable approximation of a fuzzy number should have a simple membership function, be close to the input fuzzy number, and should preserve some of its important characteristics. In this paper we suggest to approximate a fuzzy number by a piecewise linear 1-knot fuzzy number which is the closest one to the input fuzzy number among all piecewise linear 1-knot fuzzy numbers having the same core and the same support as the input. We discuss the existence of the approximation operator, show algorithms ready for the practical use and illustrate the considered concepts by examples. It turns out that such an approximation task may be problematic.

We find necessary and sufficient conditions for a ranking index defined on the set of triangular fuzzy numbers as a linear combination of its components to rank effectively. Then, based on this result, we characterize the class of ranking indices which generates orderings on triangular fuzzy numbers satisfying the basic requirements by Wang and Kerre in a slightly modified form.

Ranking fuzzy numbers is an important aspect in dealing with fuzzy optimization problems in many areas. Although so far, many fuzzy ranking methods have been discussed. This paper proposes a new Pareto approach over triangular fuzzy numbers. The approach is composed of two dominance stages. In the first stage, mono-objective dominance relations are introduced and tested with some examples. In the second stage, a Pareto dominance is defined for multi-objective optimization and then applied to solve a vehicle routing problem (VRP).

The notion of MI-group introduced in [1], [2] and later on elaborated in [3] is redefined and its structure analysed. In our approach, the role of the “Many Identities” set is replaced by an involutive anti-automorphism. Every finite MI-group coincides with some classical group, whilst infinite MI-groups comprise two parts: a group part and a semigroup part.

Uncertain data due to imprecise measurements is commonly specified as bounded interval parameters in a constraint problem. For tractability reasons, existing approaches assume independence of the parameters. This assumption is safe, but can lead to large solution spaces, and a loss of the problem structure. In this paper we propose to combine the strengths of two frameworks to tackle parameter dependency effectively, namely constraint programming and regression analysis. Our methodology is an iterative process. The core intuitive idea is to account for data dependency by solving a set of constraint models such that each model uses data parameter instances that satisfy the dependency constraints. Then we apply a regression between the parameter instances and the corresponding solutions found to yield a possible relationship function. Our findings show that this methodology exploits the strengths of both paradigms effectively, and provides valuable insights to the decision maker by accounting for parameter dependencies.

One of the major limitations to the use of brain-computer interfaces (BCIs) based on near-infrared spectroscopy (NIRS) in realistic interaction settings is the long calibration time needed before every use in order to train a subject-specific classifier. One way to reduce this calibration time is to use data collected from other users or from previous recording sessions of the same user as a training set. However, brain signals are highly variable and using heterogeneous data to train a single classifier may dramatically deteriorate classification performance. This paper proposes a transfer learning framework in which we model brain signals variability in the feature space using a bipartite graph. The partitioning of this graph into sub-graphs allows creating homogeneous groups of NIRS data sharing similar spatial distributions of explanatory variables which will be used to train multiple prediction models that accurately transfer knowledge between data sets.

Physiology-based emotionally intelligent paradigms provide an opportunity to enhance human computer interactions by continuously evoking and adapting to the user experiences in real-time. However, there are unresolved questions on how to model real-time emotionally intelligent applications through mapping of physiological patterns to users’ affective states.

In this study, we consider an approach for design of fuzzy affective agent based on the concept of typicality. We propose the use of typicality degrees of physiological patterns to construct the fuzzy rules representing the continuous transitions of user’s affective states. The approach was tested on experimental data in which physiological measures were recorded on players involved in an action game to characterize various gaming experiences. We show that, in addition to exploitation of the results to characterize users’ affective states through typicality degrees, this approach is a systematic way to automatically define fuzzy rules from experimental data for an affective agent to be used in real-time continuous assessment of user’s affective states.

Figuring out which hypothesis best explain an observed ongoing situation can be a critical issue. This paper introduces a generic model based approach to support users during this task. It then focuses on an hypothesis relevance scoring function that helps users to efficently build a convincing argumentation towards hypothesis. This function uses a multi-level extension of Yager’s aggregation algorithm, exploiting both the strength of the components of an argumentation, and the confidence the user puts in them. The presented work was illustrated on a maritime surveillance application.

The paper discusses analogical proportions in relation with the square of oppositions, a classical structure in Ancient logic which is related to the different forms of statements that may be involved in deductive syllogisms. The paper starts with a short reminder on the logical modeling of analogical proportions, viewed here as Boolean expressions expressing similarities and possibly differences between four items, as in the statement “

a

is to

b

as

c

is to

d

”. The square of oppositions and its hexagon-based extension is then restated in a knowledge representation perspective. It is observed that the four vertices of a square of oppositions form a constrained type of analogical proportion that emphasizes differences. In fact, the different patterns making an analogical proportion true can be covered by a square of oppositions or by a “square of agreement”, leading to disjunctive expressions of the analogical proportion. Besides, an “analogical octagon” is shown to capture the general construction of an analogical proportion from two sets of properties. Since the square of oppositions offers a common setting relevant for syllogisms and analogical proportions, it also provides a basis for the discussion of the possible interplay between deductive arguments and analogical arguments.

In this paper we propose a new deliberation process based on argumentation and bipolar decision making in a context of agreed common knowledge and priorities together with private preferences. This work is inspired from the supply chain management domain and more precisely by the “Collaborative Planning, Forecasting and Replenishment” model which aims at selecting a procurement plan in collaborative supply chains.

Argument graphs are a common way to model argumentative reasoning. For reasoning or computational purposes, such graphs may have to be encoded in a given logic. This paper aims at providing a systematic approach for this encoding. This approach relies upon a general, principle-based characterization of argumentation semantics.

In this paper, we will first discuss fuzzy generalized quantifiers and their formalization in the higher-order fuzzy logic (the fuzzy type theory). Then we will briefly introduce a special model of intermediate quantifiers, classify them as generalized fuzzy ones and prove that they have the general properties of isomorphism invariance, extensionality and conservativity. These properties are characteristic for the quantifiers of natural language.

The drastic product *

D

is known to be the smallest

t

-norm, since

x

*

D

y

 = 0 whenever

x

,

y

 < 1. This

t

-norm is not left-continuous, and hence it does not admit a residuum. So, there are no drastic product

t

-norm based many-valued logics, in the sense of [7]. However, if we renounce standard completeness, we can study the logic whose semantics is provided by those MTL chains whose monoidal operation is the drastic product. This logic is called S

3

MTL in [17]. In this note we justify the study of this logic, which we rechristen DP (for drastic product), by means of some interesting properties relating DP and its algebraic semantics to a weakened law of excluded middle, to the Δ projection operator and to discriminator varieties. We shall show that the category of finite DP-algebras is dually equivalent to a category whose objects are multisets of finite chains. This duality allows us to classify all axiomatic extensions of DP, and to compute the free finitely generated DP-algebras.

The urban-type residential houses built before World War Two represent a large part of the built environment in Hungary. Due to their physical condition and low energy-efficiency the retrofit of these buildings is very much advisable nowadays. In this paper we propose an approach based on fuzzy signatures and state machines, that helps decision support for determining the renovation scenario concerning necessity, cost efficiency and quality. Using the knowledge obtained from diagnostic surveys done during the previous decades by architect experts, and technical guides and the available database of contractors billing, a protocol for the preparation for optimized refurbishment is proposed, based on the concept of an extended fuzzy state machine model. In this combined model the theoretical concepts of finite-state machine and fuzzy state machine, and also the principles of fuzzy signatures are applied.

In this paper we use tools from the theory of MV-algebras and MV-algebraic states to study infinitesimal perturbations of classical (i.e. Boolean) events and their non-Archimedean probability. In particular we deal with a class of MV-algebras which can be roughly defined by attaching a cloud of infinitesimals to every element of a finite Boolean algebra and for them we introduce the class of Chang-states. These are non-Archimedean mappings which we prove to be representable in terms of a usual (i.e. Archimedean) probability measure and a positive group homomorphism capable to handle the infinitesimal side of the MV-algebras we are dealing with. We also study in which relation Chang-states are with MV-homomorphisms taking value in a suitable perfect MV-algebra.

Gradual itemsets of the form “

the more/less A, the more/less B

” summarize data through the description of their internal tendencies, identified as correlation between attribute values. This paper proposes to enrich such gradual itemsets by taking into account an acceleration effect, leading to a new type of gradual itemset of the form “

the more/less A increases, the more quickly B increases

”. It proposes an interpretation as convexity constraint imposed on the relation between

A

and

B

and a formalization of these accelerated gradual itemsets, as well as evaluation criteria. It illustrates the relevance of the proposed approach on real data.

In this paper we propose a new type of protoform-based linguistic summary – the gradual summary. This new type of summaries aims in capturing the change over some time span. Such summaries can be useful in many domains, for instance in economics, e.g., “prices of X are getting smaller”, in eldercare, e.g., “resident Y is getting less active”, in managing production, e.g. “production is dropping” or “delays in deliveries are getting smaller”.

Dengue fever is the world’s fastest growing vector-borne disease. Studying such data aims at better understanding the behaviour of this disease to prevent the dengue propagation. For instance, it may be the case that the number of cases of dengue fever in cities depends on many factors, such as climate conditions, density, sanitary conditions. Experts are interested in using geographical information systems in order to visualize knowledge on maps. For this purpose, we propose to build maps based on gradual patterns. Such maps provide a solution for visualizing for instance the cities that follow or not gradual patterns.

We present a novel comprehensive model of a consensus reaching support system in the fuzzy context. We assume the individual fuzzy preferences, a fuzzy majority in group decision making, as proposed by Kacprzyk [9], some fuzzy majority based solution concepts in group decision making, notably fuzzy cores (cf. Kacprzyk [9]) and their choice function based representations by Kacprzyk and Zadrożny [15],[16], a soft degree of consensus by Kacprzyk and Fedrizzi [10],[11]. Using as a point of departure Kacprzyk and Zadrożny’s [18] approach of the use of linguistic data summaries to support the running of a consensus reaching process, we develop and implement a novel approach that synergistically combines the tools and techniques mentioned above. We assume that moderated consensus reaching process which is run in the group of agents by a special agent called a moderator, is the most effective and efficient solution. We attempt to facilitate the work of a moderator, by some useful guidelines and additional indicators. We extend this idea and finally, we present a new implementation followed by a numerical evaluation of the new model proposed.

We consider aggregation of partially known qualitative preferences for a group of agents, considering necessary and potentially optimal choices with respect to different notions of optimality (consensus, extreme choices, Pareto optimality) and provide a theoretical characterization. We report statistics (obtained with simulations with synthetic data) about the cardinality of the sets of possible and necessarily optimal choices for the different cases. Finally we introduce preliminary ideas on a qualitative notion of

fairness

and on interactive elicitation.

The classical von Neumann–Morgenstern’s notion of lottery is generalized by replacing a probability distribution on a finite support with a belief function on the power set of the support. Given a partial preference relation on a finite set of generalized lotteries, a necessary and sufficient condition (weak rationality) is provided for its representation as a Choquet expected utility of a strictly increasing utility function.

With the growing interest in Multi-Agent Systems (MAS) based solutions, one can find multiple MAS conceptions and implementations dedicated to the same goal. Those systems with their complex behaviors are rarely predictable. They may provide different results according to agents’ interactions sequences. Consequently, evaluation of the quality of MAS returned results became an urgent need. Our approach is interested in evaluating high level data by considering agent’s preferences regarding performatives. By analogy with the economic field, agents may ask for services, so they are

consumers

and may receive different possible answers to their requests from other agents which are

producers

. We will then focus on the analysis of messages exchanged within standard interaction protocols and compute the utility value associated to every conversation. Then we conclude utility measures for each agent and for the whole MAS regarding some execution results.

This work focuses on label ranking, a particular task of preference learning, wherein the problem is to learn a mapping from instances to rankings over a finite set of labels. This paper discusses and proposes alternative reduction techniques that decompose the original problem into binary classification related to pairs of labels and that can take into account label correlation during the learning process.

In this work we propose a clustering methodology model named as Mixture of Fuzzy Models (MFMs). We adopt two assumptions: the data points are generated by a membership function and the sum of the memberships to all of the clusters must be greater or equal than zero. The objective is to obtain a set of membership functions which represent the data. It is formulated as a multiobjective optimization problem with two objectives: to maximize the sum of memberships within each cluster and to maximize the differences of memberships between clusters.

Analogical proportion-based classification methods have been introduced a few years ago. They look in the training set for suitable triples of examples that are in an analogical proportion with the item to be classified, on a maximal set of attributes. This can be viewed as a lazy classification technique since, like k-nn algorithms, there is no static model built from the set of examples. The amazing results (at least in terms of accuracy) that have been obtained from such techniques are not easy to justify from a theoretical viewpoint. In this paper, we show that there exists an alternative method to build analogical proportion-based learners by statically building a set of inference rules during a preliminary training step. This gives birth to a new classification algorithm that deals with pairs rather than with triples of examples. Experiments on classical benchmarks of the UC Irvine repository are reported, showing that we get comparable results.

In this paper, we study how multilabel predictions can be obtained when our uncertainty is described by a convex set of probabilities. Such predictions, typically consisting of a set of potentially optimal decisions, are hard to make in large decision spaces such as the one considered in multilabel problems. However, we show that when considering the Hamming loss, an approximate prediction can be efficiently computed from label-wise information, as in the precise case. We also perform some first experiments showing the interest of performing partial predictions in the multilabel case.

A safe and reliable control operation can be difficult due to limitations in operator’s skills. A self-developing control system could help assist or even replaces the operators in providing the required control operations. However, the self-developing control system is lack of flexibility in determining the necessary control option in multiple conditions where a human operator usually prevails by experiences in optimizing priority. Here, a cooperative multi-knowledge learning control system is proposed in providing flexibility for determining priority in control options, within multiple conditions by considering the required self-developing control knowledge in fulfilling these conditions. The results show that the system was able to provide consideration in prioritizing the use of the required control knowledge of the condition assigned.

Especially for one-class classification problems, an accurate model of the class is necessary. Since the shape of a class might be arbitrarily complex, it is hard to choose an approach that is generic enough to cope with the variety of shapes, while delivering an interpretable model that remains as simple as possible and thus applicable in practice. In this article, this problem is tackled by combining convex building blocks both additively and subtractively in a tree-like structure. The convex building blocks are represented by multivariate membership functions that aggregate the respective parts of the learning data. During the learning process, proven methods from support vector machines and cluster analysis are employed in order to optimally find the structure of the tree. Several academic examples demonstrate the viability of the approach.

In figuring out the complete content of a fictional story, all kinds of consequences are drawn from the explicitly given material. It may seem natural to assume a closure deductive principle for those consequences. Notwithstanding, the classical closure principle has notorious problems because of the possibility of inconsistencies. This paper aims to explore an alternative approach to reasoning with the content of fictional works, based on the application of a mathematical model for conjectures, hypotheses and consequences (abbr. CHCs), extensively developed during the last years by Enric Trillas and some collaborators, with which deduction in this setting becomes more comprehensive.

Zadeh posed several Computing with Words (CWW) test problems such as: “What is the probability that John is short?” These problems assume a given piece of information in the form of membership functions for linguistic terms including tall, short, young, middle-aged, and the probability density functions of age and height. This paper proposes a solution that interprets Zadeh’s solution for these problems as a solution in terms of probability spaces as defined in the probability theory. This paper also discusses methodological issues of relations between concepts of probability and fuzzy sets.

The aim of this paper is to open a critical discussion on the claim, recently presented in the community and especially heralded by Enric Trillas, that fuzzy logic should be seen as an “experimental science”. The first interesting aspect of such remark is whether and in which way such position has consequences on the real development of the research, or if it is simply a (different) way of looking at the same phenomenon. As a consequence, we investigate the possible connection to Zadeh’s distiction between Fuzzy logic in a restricted sense and in a general sense. We shall argue that Trillas’s claim not only strongly supports the necessity for such a distinction, but provides a path of investigation which can preserve the conceptual innovativeness of the notion of fuzziness.

How do humans develop concepts? – This paper presents a historical view on answers to this question. Psychologist Piaget was influenced by Philosopher Kant when he founded his theory of cognitive child development named “Genetic Epistemology”. Biologist and historian of science Rheinberger emphasized that scientific concepts are “fluctuating objects” or “imprecise concepts” when he founded his “Historical Epistemology”. In this paper we combine these approaches with that of fuzzy concepts. We give some hints to establish a new approach to extend Piaget’s theory, to a so-called “Fuzzy Genetic Epistemology”.

The evidence coming from cognitive psychology and linguistics shows that pairs of reference concepts (as e.g.

good

/

bad

,

tall

/

short

,

nice

/

ugly

, etc.) play a crucial role in the way we everyday use and understand natural languages in order to analyze reality and make decisions. Different situations and problems require different pairs of landmark concepts, since they provide the referential semantics in which the available information is understood accordingly to our goals in each context. In this way, a semantic valuation structure or system

emerges

from a pair of reference concepts and the way they oppose each other. Such structures allow representing the logic of new concepts according to the semantics of the references. We will refer to these semantic valuation structures as

paired structures

. Our point is that the semantic features of a paired structure could essentially depend on the semantic relationships holding between the pair of reference concepts from which the valuation structure emerges. Different relationships may enable the representation of different types of

neutrality

, understood here as an epistemic hesitation regarding the references. However, the standard approach to natural languages through logical models usually assumes that reference concepts are just each other complement. In this paper, we informally discuss more deeply about these issues, claiming in a positional manner that an adequate logical study and representation of the features and complexity of natural languages requires to consider more general semantic relationships between references.

Currently, tactile surfaces implemented in automobiles are passive, i.e.,

feedbackless

, thus forcing the user to visually check the device. To improve drivers’ interaction, surface vibrations can be used to deliver feedback to the finger when touched, and an associated perception model is required. Hence, this paper introduces a fuzzy model for the comfort degree of vibrotactile signals. System input variables are chosen from the physical characteristics of the signals, and are validated on a dissimilarity judgment task. The system achieves an error of 9% and correctly classifies 17 out of 18 signals within a reasonable interval. A graphical user interface to interact with the system is also presented.

In this paper, we introduce a multi-criteria and multi-expert decision-making procedure for dealing with sensory analysis. Experts evaluate each product, taking different criteria into account. If they are confident in their opinions, the evaluation is presented using specific linguistic terms. If not, linguistic expressions generated from several consecutive linguistic terms are used. Products are ranked according to the average distance between the obtained ratings and the highest possible assessment. The procedure is applied to a field experiment in which six trained sensory panelists assessed a variety of wines and wild mushrooms.

The design of automated systems for the recognition of specific human activities is among the most promising research activities in Ambient Intelligence. The literature suggests the adoption of wearable devices, relying on acceleration information to model the activities of interest and distance metrics for the comparison of such models with the run-time data. Most current solutions do not explicitly model the uncertainty associated with the recognition, but rely on crisp thresholds and comparisons which introduce brittleness and inaccuracy in the system. We propose a framework for the recognition of simple activities in which recognition uncertainty is modelled using possibility distributions. We show that reasoning about this explicitly modelled uncertainty leads to a system with enhanced recognition accuracy and precision.

3D virtual garment design using specific computer-aided-design software has attracted a great attention of textile/garment companies. However, there generally exists a perceptual gap between virtual and real products for both designers and consumers. This paper aims at quantitatively charactering human perception on virtual fabrics and its relation with the technical parameters of real fabrics. For this purpose, two sensory experiments are carried out on a small number of fabric samples. By learning from the identified input (technical parameters of the software) and output (sensory descriptors) data, we set up a series of models using different techniques. The fuzzy ID3 decision tree model has shown better performance than the other ones.

This paper presents a new methodology aimed at making simpler the product/market fit process. We propose a user-centered approach inspired on the Oriental philosophy that is behind Kansei Engineering. In essence, we advocate for customization of products guided by users’ expectations. Our proposal combines Sensory Analysis and Soft Computing techniques in order to uncover what users think but also what they feel and desire when facing new products. That is elicitation of the so-called

kanseis

or “psychological feelings”. Then, we can design new prototypes that truly matter to people because they fit the deepest users’ demands. Thus, improving innovation and marketing success rate. We have illustrated the details of our proposal in a case study related to gin packaging.