Lower probability models for uncertainty and nondeterministic processes

Patrick Suppes’s vigorous defense of the centrality of probability and of the irreducibility of randomness or chance phenomena, especially in the province of quantum mechanical phenomena, his interests in axiomatization, and the viewpoints he developed through his fundamental work on measurement theory have encouraged and enabled him to study the interpretations and formal structure of probability and to deviate from conventional probability as axiomatized by Kolmogorov. With Mario Zanotti he has pursued the concept of upper and lower or interval-valued probability initially through a measurement-theoretic algebraic viewpoint on subjective probability and subsequently through its use in meeting the challenges to probability posed by quantum mechanics. We note our parallel initial efforts to develop a concept of upper and lower probability that can be used to represent comparative probability relations and our subsequent efforts to broaden the scope of probabilistic reasoning about empirical phenomena by modeling time-invariant, nondeterministic phenomena whose long-run time averages are either divergent or have behavior about which we are ignorant.

Incremental conditioning of lower and upper probabilities

Reliable deterioration models for forecasting the condition of bridge elements, systems, and networks are essential components of any bridge management system. Estimating deterioration transition probabilities is the basis of the popular Markov chain deterioration model. Transition probabilities ha ve been produced in the literature by statistical and probabilistic analysis performed on available bridge inspection and condition data. The available data may be limited or inconsistent with the Markov chain transition period. In addition, the conventional probability theory has limitations when applied to problems with a stochastic nature such as bridge deterioration. This paper introduces a novel method based on the theory of evidence for bridge deterioration modeling through expert judgment elicitation. The advantages of the theory of evidence over the traditional probability theory are discussed and the process for the theory implementation is demonstrated with a case study to validate the application of Dempster–Shafer theory of evidence to estimate the transition probabilities. Based on the results, the theory of evidence is proposed as a scientific expert judgement elicitation technique in the area of bridge condition rating and deterioration modeling. Expert judgment elicitation and theory of evidence application hold potential in the field of bridge management and require further investigation and research.

The paper proposes a parallel survey of revision and updating operations available in the probability theory and in the possibility theory frameworks. In these two formalisms the current state of knowledge is generally represented by a [0,1]-valued function whose domain is an exhaustive set of mutually exclusive possible states of the world. However, in possibility theory, the unit-interval can be viewed as a purely ordinal scale. Two general kinds of operations can be defined on this assignment function: conditioning, and imaging (or “projection”). the difference between these two operations is analogous to the one made between belief revision à la Gärdenfors and updating à la Katsuno and Mendelzon in the logical framework. In the probabilistic framework these two operations are respectively Bayesian conditioning and Lewis' imaging. Counterparts to these operations are presented for the possibilistic framework including the case of conditioning upon uncertain observations, and justifications are given which parallel the ones existing for the probabilistic operations. More particularly, it is recalled that possibilistic conditioning satisfies all the postulates proposed by Alchourrón, Gärdenfors and Makinson for belief revision (stated in possibilistic terms), and it is proved that possibilistic imaging satisfies all the postulates proposed by Katsuno and Mendelzon. the situation where our current knowledge is stated in terms of weighted logical propositions is discussed in connection to possibility theory. Revision in other more complex numerical formalisms, namely belief and plausibility functions, and upper and lower probabilities is also surveyed. Recent results on the revision of conditional knowledge bases are also reviewed. the frameworks of belief functions, upper and lower probabilities and conditional bases are more sophisticated than the previous ones because they enable to distinguish between factual evidence and generic knowledge in a cognitive state. This, framework leads to two forms of belief revision respectively taking care of the revision of evidence and the revision of knowledge. © 1994 John Wiley & Sons, Inc.

ABSTRACTThe main contribution of this paper is a new definition of expected value of belief functions in the Dempster–Shafer (D–S) theory of evidence. Our definition shares many of the properties of the expectation operator in probability theory. Also, for Bayesian belief functions, our definition provides the same expected value as the probabilistic expectation operator. A traditional method of computing expected of real-valued functions is to first transform a D–S belief function to a corresponding probability mass function, and then use the expectation operator for probability mass functions. Transforming a belief function to a probability function involves loss of information. Our expectation operator works directly with D–S belief functions. Another definition is using Choquet integration, which assumes belief functions are credal sets, i.e. convex sets of probability mass functions. Credal sets semantics are incompatible with Dempster's combination rule, the center-piece of the D–S theory. In general, our definition provides different expected values than, e.g. if we use probabilistic expectation using the pignistic transform or the plausibility transform of a belief function. Using our definition of expectation, we provide new definitions of variance, covariance, correlation, and other higher moments and describe their properties.

Artificial Intelligence has emerged and been considerably developed in the last fifty years together with the advent of the computer age, as a new scientific area aiming at processing information in agreement with the way humans do. Inevitably, such a range of efforts, both theoretical and application-oriented, and motivated by new concerns, has led to modify and enlarge the way basic notions such as uncertainty, belief, knowledge, or evidence could be thought, represented and handled in practice. In particular, there has been a major trend in uncertainty (more specifically, partial belief) modelling, emphasizing the idea that the degree of confidence in an event is not totally determined by the confidence in the opposite event, as assumed in probability theory. Possibility theory (Dubois and Prade, 1988) belongs to this trend that describes partial belief in terms of certainty and plausibility, viewed as distinct concepts. Belief and plausibility functions (Shafer, 1976), or lower and upper probabilities (Walley, 1991) are other important representatives of this trend. The distinctive features of possibility theory are its computational simplicity, and its position as a bridge between numerical and symbolic theories of partial belief for practical reasoning. Possibility theory, based on a non-additive setting that contrasts with probability theory, provides a potentially more qualitative treatment of partial belief, since the operations “max” and “min” play a role somewhat analogous to the sum and the product in probability calculus. There are two different kinds of possibility theory: one is qualitative and the other is quantitative. They share the same kind of set-functions, but they differ when it comes to conditioning and combination

Since its introduction in the 1960s, Dempster-Shafer theory became one of the leading strands of research in artificial intelligence with a wide range of applications in business, finance, engineering and medical diagnosis. In this paper, we aim to grasp the essence of Dempster-Shafer theory by distinguishing between ambiguous-and-questionable and ambiguous-but-correct perceptions. Throughout the paper, we reflect our analysis in terms of signals and sensors as a natural field of application. We model ambiguous-and-questionable perceptions as a probability space with a quantity random variable and an additional perception random variable (Dempster model). We introduce a correctness property for perceptions. We use this property as an axiom for ambiguous-but-correct perceptions. In our axiomatization, Dempster’s lower and upper probabilities do not have to be postulated: they are consequences of the perception correctness property. Even more, we outline how Dempster’s lower and upper probabilities can be understood as best possible estimates of quantity probabilities. Finally, we define a natural knowledge fusion operator for perceptions and compare it with Dempster’s rule of combination.

<div class="htmlview paragraph">This paper investigates the potential of Evidence Theory (ET) and Bayesian Theory (BT) for decision under uncertainty, when the evidence about uncertainty is imprecise. The basic concepts of ET and BT are introduced and the ways these theories model uncertainties, propagate them through systems and assess the safety of these systems are presented. ET and BT approaches are demonstrated and compared on examples involving an algebraic function when the evidence about the input variables consists of intervals provided by experts. It is recommended that a decision maker compute both the Bayesian probability of events and their lower and upper probabilities using ET when evidence from experts is imprecise. A large gap between the lower and upper probability suggests that more information should be collected before making a decision. If this is not feasible, then Bayesian probabilities can help make a decision.</div>

As is well known, the Dempster-Shafer mathematical theory of evidence, which employs the use of belief functions, has been used to perform uncertain inference in expert systems and artificial intelligence. As is also well known, many fault this discipline for two reasons: (1) Dempster-Shafer theory is not probability theory, and there exist cogent arguments for using only probability theory to perform uncertain inference; (2) As shown by a number of authors, mechanical applications of belief functions can lead to quite unacceptable results. In a 1990 special issue of the International Journal of Approximate Reasoning, which was devoted to a discussion of belief functions, these criticisms were forwarded and various interpretations of belief functions were offered. In his response to the discussion in the special issue, Shafer showed the problems with these criticisms and interpretations, and he once again explained how he meant for belief functions to be interpreted. In this article, this defense is furthered in two ways: (1) It is shown that belief functions, as Shafer intends them to be interpreted, use probability theory in the same way as the traditional statistical tool, significance testing; and (2) A problem is given for which an application of belief functions yields a meaningful solution while a Bayesian analysis does not.

Mathematical theories of uncertainty in which measures of uncertainty are now well established are fuzzy set theory, evidence theory, and possibility theory, in addition to classical set theory and probability theory. Some of these theories are connected. Classical set theory is subsumed under fuzzy set theory. Probability theory and possibility theory are branches of evidence theory, while evidence theory is, in turn, a branch of fuzzy measure theory. Although information aspects of fuzzy measure theory have not been investigated as yet, the theory is briefly introduced in this chapter because it represents a broad framework for future research.

Whereas probability theory has been very successful as a conceptual framework for risk analysis in many areas where a lot of experimental data and expert knowledge are available, it presents certain limitations in applications where only weak information can be obtained. One such application investigated in this paper is water treatment, a domain in which key information such as input water characteristics and failure rates of various chemical processes is often lacking. An approach to handle such problems is proposed, based on the Dempster-Shafer theory of belief functions. Belief functions are used to describe expert knowledge of treatment process efficiency, failure rates, and latency times, as well as statistical data regarding input water quality. Evidential reasoning provides mechanisms to combine this information and assess the plausibility of various noncompliance scenarios. This methodology is shown to boil down to the probabilistic one where data of sufficient quality are available. This case study shows that belief function theory may be considered as a valuable framework for risk analysis studies in ill-structured or poorly informed application domains

The aim of this paper is to show that the Kantorovich problem, well known in models of economics and very intensively studied in probability theory in recent years, can be viewed as the basis of some constructions in the theory of belief functions. We demonstrate this by analyzing specialization relation for finitely defined belief functions and belief functions defined on reals. In addition, for such belief functions we consider the Wasserstein metric and study its connections to disjunctions of belief functions.

Dempster-Shafer evidence theory, as a generalization of probability theory, is a powerful tool for dealing with a variety of uncertainties, such as incompleteness, ambiguity, and conflict. Because of its advantages in information fusion compared with traditional probability theory, it is widely used in various fields. However, the classic Dempster's combination rule assumes that evidences are independent of each other, which is difficult to satisfy in real life. Ignoring the dependence among the evidences will lead to unreasonable fusion results, and even wrong conclusions. Considering the limitations of D-S evidence theory, this paper proposed a new evidence fusion model based on principal component analysis (PCA) to deal with the dependence among evidences. First, the approximate independent principal components of each information source were obtained based on principal component analysis. Second, the principal component data set was used as a new information source for evidence theory. Third, the basic belief assignments (BBAs) were constructed. As the fundamental construct of evidence theory, a BBA is a probabilistic function corresponding to each hypothesis, quantifying the belief assigned based on the evidence at hand. This function facilitates the synthesis of disparate evidence sources into a mathematically coherent and unified belief structure. After constructing the BBAs, the BBAs were fused and a conclusion was drawn. The case study verified that the proposed method is more robust than several traditional methods and can deal with redundant information effectively to obtain more stable results.

Uncertainty measure in evidence theory supplies a new criterion to assess the quality and quantity of knowledge conveyed by belief structures. As generalizations of uncertainty measure in the probabilistic framework, several uncertainty measures for belief structures have been developed. Among them, aggregate uncertainty AU and the ambiguity measure AM are well known. However, the inconsistency between evidential and probabilistic frameworks causes limitations to existing measures. They are quite insensitive to the change of belief functions. In this paper, we consider the definition of a novel uncertainty measure for belief structures based on belief intervals. Based on the relation between evidence theory and probability theory, belief structures are transformed to belief intervals on singleton subsets, with the belief function Bel and the plausibility function Pl as its lower and upper bounds, respectively. An uncertainty measure SU for belief structures is then defined based on interval probabilities in the framework of evidence theory, without changing the theoretical frameworks. The center and the span of the interval is used to define the total uncertainty degree of the belief structure. It is proved that SU is identical to Shannon entropy and AM for Bayesian belief structures. Moreover, the proposed uncertainty measure has a wider range determined by the cardinality of discernment frame, which is more practical. Numerical examples, applications and related analyses are provided to verify the rationality of our new measure.

Fatigue failure is the main failure mode of mechanical components in the research of engineering structures. As fatigue life may be a basis for the fatigue reliability design, it is very important to predict it for the normal usage of the structure. Uncertainties rooted in physical variability, data uncertainty and modeling errors of the fatigue life prediction analysis. Furthermore, the predicted life of concrete structures in civil engineering field will be more obviously uncertain than other engineering structures. Due to lack of knowledge or incomplete, inaccurate, unclear information in the modeling, there are limitations in using only one framework (probability theory) to quantify the uncertainty in the concrete fatigue life prediction problem because of the impreciseness of data or knowledge. Therefore the study of uncertainty theory in the prediction of fatigue life is very necessary. This study explores the use of evidence theory for concrete fatigue life prediction analysis in the presence of epistemic uncertainty. The empirical formula S-N curve and the Paris law based on the fracture mechanics are selected as the fatigue life prediction models. The evidence theory is used to quantify the uncertainty present in the models' parameters. The parameters in fatigue damage prognosis model are obtained by fitting the available sparse experimental data and then the uncertainty in these parameters is taken into account. In order to alleviate the computational difficulties in the evidence theory based uncertainty quantification (UQ) analysis, a differential evolution (DE) based interval optimization method is used for finding the propagated belief structure. The object of the current study is to investigate uncertainty of concrete fatigue damage prognosis using sparse experimental data in order to explore the feasibility of the approach. The proposed approach is demonstrated using the experimental results of the plain concrete beams and the steel fibred reinforced concrete beams.