Based on Klir’s framework of uncertainty, the total uncertainty (also called ambiguity) of belief function is linear addition of discord and nonspecificity. Though uncertainty measures of belief function have been discussed widely, there is no measure that can satisfy the monotonicity and range consistency properties at the same time. In this article, we discuss uncertainty measure of belief function from the perspective of information fractal dimension. An uncertainty quantity called evenness and its measure ${\rm Eve}$ are proposed, which can represent the belief propensity degree of belief function. We first propose the measures of diversity (normalized nonspecificity) and the element evenness (normalized discord), and then fuse them to calculate ${\rm Eve}$ . The proposed method can not only measure the subnormal mass function but also interpret the different views of Klir and Smets on “Uncertainty.” In addition, we extend Klir’s framework of uncertainty based on the proposed information quantities.
21 November 2017 in IEEE Transactions on Systems, Man, and Cybernetics: Systems
The theory of evidence (TE) has been largely used for many applications. This theory is a generalization of probability distribution and offers a mathematical representation for two types of uncertainty-based information: 1) discord and 2) nonspecificity. Several measures have already been developed to quantify these two types of uncertainty. They have been called total uncertainty (TU) measures since they quantify both types of uncertainty. The generalized Hartley measure and the maximum entropy have been the only measures so far that satisfy a list of properties very desirable for practical applications. Recently, two new measures of nonspecificity and TU based on belief intervals have been proposed. These two measures do not satisfy the properties of additivity, superadditivity, and subadditivity in the TE. The present critique is about these shortcomings and provides a more complete analysis of those uncertainty measures with respect to a list of desired properties. A potential consequence of an ill-characterized measure may yield selecting an inappropriate rule for decision-making in the processing chain from data to information to decisions.
9 June 2016 in IEEE Transactions on Image Processing
This paper proposes an approach referred as: iterative refinement of possibility distributions by learning (IRPDL) for pixel-based image classification. The IRPDL approach is based on the use of possibilistic reasoning concepts exploiting expert knowledge sources as well as ground possibilistic seeds learning. The set of seeds is constructed by incrementally updating and refining the possibility distributions. Synthetic images as well as real images from the RIDER Breast MRI database are being used to evaluate the IRPDL performance. Its performance is compared with three relevant reference methods: region growing, semi-supervised fuzzy pattern matching, and Markov random fields. The IRDPL performance (in terms of recognition rate, 87.3%) is close to the Markovian method (88.8%) that is considered to be the reference in pixel-based image classification. IRPDL outperforms the other two methods, respectively, at the recognition rates of 83.9% and 84.7%. In addition, the proposed IRPDL requires fewer parameters for the mathematical representation and presents a reduced computational complexity.
Uncertainty is at the heart of decision-making processes in most real-world applications. Uncertainty can be broadly categorized into two types: aleatory and epistemic. Aleatory uncertainty describes the variability in the physical system where sensors provide information (hard) of a probabilistic type. Epistemic uncertainty appears when the information is incomplete or vague such as judgments or human expert appreciations in linguistic form. Linguistic information (soft) typically introduces a possibilistic type of uncertainty. This paper is concerned with the problem of classification where the available information, concerning the observed features, may be of a probabilistic nature for some features, and of a possibilistic nature for some others. In this configuration, most encountered studies transform one of the two information types into the other form, and then apply either classical Bayesian-based or possibilistic-based decision-making criteria. In this paper, a new hybrid decision-making scheme is proposed for classification when hard and soft information sources are present. A new Possibilistic Maximum Likelihood (PML) criterion is introduced to improve classification rates compared to a classical approach using only information from hard sources. The proposed PML allows to jointly exploit both probabilistic and possibilistic sources within the same probabilistic decision-making framework, without imposing to convert the possibilistic sources into probabilistic ones, and vice versa.
Measuring similarity is of a great interest in many research areas such as in data sciences, machine learning, pattern recognition, text analysis and information retrieval to name a few. Literature has shown that possibility is an attractive notion in the context of distinguishability assessment and can lead to very efficient and computationally inexpensive learning schemes. This paper focuses on determining the similarity between two possibility distributions. A review of existing similarity measures within the possibilistic framework is presented first. Then, similarity measures are analyzed with respect to their capacity to satisfy a set of required properties that a similarity measure should own. Most of the existing possibilistic similarity measures produce undesirable outcomes since they generally depend on the application context.A new similarity measure, called InfoSpecificity, is introduced and the similarity measures are categorized into three main methods: morphic-based, amorphic-based and hybrid. Two experiments are being conducted using four benchmark databases. The aim of the experiments is to compare the efficiency of the possibilistic similarity measures when applied to real data. Empirical experiments have shown good results for the hybrid methods, particularly with the InfoSpecificity measure. In general, the hybrid methods outperform the other two categories when evaluated on small-size samples, i.e., poor-data context (or poor-informed environment) where possibility theory can be used at the greatest benefit.