Certainty Distributions Research Articles

Rainbow UDA: Combining Domain Adaptive Models for Semantic Segmentation Tasks.

In this paper, we propose Rainbow UDA, a framework designed to address the drawbacks of the previous ensemble-distillation frameworks when combining multiple unsupervised domain adaptation (UDA) models for semantic segmentation tasks. Such drawbacks are mainly resulted from overlooking the magnitudes of the output certainties of different members in an ensemble as well as their individual performances in the target domain, causing the distillation process to suffer from certainty inconsistency and performance variation issues. These issues may hinder the effectiveness of an ensemble that includes members with either biased certainty distributions or have poor performance in the target domain. To mitigate such a deficiency, Rainbow UDA introduces two operations: the unification and the channel-wise fusion operations, to address the above two issues. In order to validate the designs of Rainbow UDA, we leverage the GTA5 →Cityscapes and SYNTHIA → Cityscapes benchmarks to examine the effectiveness of the two operations, and compare Rainbow UDA against a wide variety of baseline approaches. We also provide a set of analyses to show that Rainbow UDA is effective, robust, and can evolve with time as the ensemble grows.

Meta-strategies of referencing basic level categories in elaborative response generation in cognitive agents

This paper introduces a set of meta-strategies utilising basic-level categories in a cognitive agent capable of answering question using autoepistemic natural statements grounded in its experience and hierarchical knowledge structures. The proposed approach thrives in combing several natural phenomena, as observed in human language users, with inherent notion of uncertainty about categorisation of observable objects. Using built-in hierarchical structures between available concepts we show how a set of basic meta-strategies can be defined to determine the focus of the response (responding agent) in the context of particular distribution of certainty (established through past individual experiences) in the hierarchy, current focus of the question (querying agent) and basic-level in the hierarchy. Such meta-strategies can be further implemented as more natural extensions of agent's internal organisation of linguistic activities.

Investigation of epistemological indicators in socio-political talk shows of Iran TV from the perspective of critical discourse [In Persian

Learning analysis is performed using meta-discourse markers. Learning markers are, in fact, linguistic forms that are considered as important rhetorical tools and ensure interaction in texts. Different models for learning analysis are presented. the Highland pattern is one of the most common and popular. For the first time, Highland has defined and categorized all aspects of the interaction between the author of the text and the wisdom of the text, as well as the author of the text and the reader of the text, using the term learning, and claims that the researcher can go beyond text and discourse and also analyze the text as a social action. Learning is used as a concept in discourse analysis and is used in the interaction between authors and their text between writer and reader. Overall, learning, as an important tool in facilitating communication, helps the writer to communicate with drivers and express his personal findings, as propositions place their relevant sentences or phrases in creating a coherent discourse in the field of text structure. The present article examines the epistemological indicators in socio-political talk shows on the Islamic Republic of Iran TV channels from the perspective of critical discourse based on the views of Highland (2005) and Fairclough (2001). In this research, 20 programs of socio-political talk show of channels 1, 2, 3 and 4 were selected, studied and analyzed qualitatively and quantitatively in a numerical manner using purposive sampling. The variables studied in this study include the gender of the experts, whether the programs are live or recorded, the number of experts in the program, whether they are academics or non-academics to determine the results showed that the distribution of epistemological elements, certainty, skepticism and attitude In socio-political roundtable programs on the Islamic Republic of Iran Broadcasting, it varies according to the gender of the experts, whether the program is live or recorded, whether the experts are academic or non-academic, or the presence of one or more experts in the program. The type of application of epistemological tools in the discourse of experts can be justified.

Source attribution at the food sub-product level for the development of the Canadian Food Inspection Agency risk assessment model

Decreasing the health burden caused by foodborne pathogens is challenging and it depends on the identification of the most significant hazards and food sources causing illnesses, so adequate mitigation strategies can be implemented. In this regard, the Canadian Food Inspection Agency (CFIA) has developed the Establishment-based Risk Assessment (ERA) model, so that a more effective and efficient allocation of resources can be assigned to the highest food safety risk areas. To assess risk, the model considers the type of food sub-products being manufactured by establishments and its scope is limited to the 17 most important foodborne pathogens representing the highest level of food safety risk. However, the information on source attribution at the sub-product level based on a structured approach is limited. To overcome this challenge, an expert elicitation was conducted in 2016 to estimate the relative contribution and associated certainty of each sub-product for 31 pathogen-commodity combinations to the total Canadian health burden associated with foodborne illnesses (expressed in DALYs). These DALYs represent 78% of the total Canadian health burden associated with federally-regulated food commodities considered within the model. A total of 49 Canadian experts recruited using a “snow ball” sampling strategy participated in the study by completing an electronic survey. Results of the elicitation displayed variable levels of health burden allocation between the pathogens and the different commodity sub-products. Assessment of the certainty levels showed some combinations being evaluated with more confidence (e.g., Campylobacter and eggs/poultry sub-products) than others, where a bimodal distribution of certainty was observed (e.g., Toxoplasma in pork sub-products). Furthermore, no participant raised concerns on the food classification scheme, suggesting their agreement with the proposed sub-products categorization of the elicitation. Relative contribution estimates will be included in the CFIA ERA model and used to enhance its applicability for risk prioritization and effective resource allocation during food establishment inspections. While substantial uncertainty around the central tendency estimates was found, these estimates provide a good basis for regulatory oversight and public health policy.

On Backlash: Emotion and the Politicisation of Security

This article explores the role of emotion in the politicisation of security through the concept of backlash: the idea of visceral and reactionary episodes where security claims are adamantly rejected and the subject of ‘security’ becomes intensely controversial. Starting by examining the role of emotion in politicisation, I make the case for viewing emotions as playing a key role in the distribution of certainty in security discourse. Building on this epistemic view of emotion, I review how backlash is understood in other fields before tailoring a definition for security studies centered around four constitutive features: reaction, hostility, emotion, and contagion. The final section focuses on the politicising effects of backlash including the mobilisation of backlash movements, the intensification of controversy, and arena shifting. The discussion concludes by suggesting that the concept of backlash offers a promising research agenda for those inquiring into the politicisation of security.

Recognizing input space and target concept drifts in data streams with scarcely labeled and unlabelled instances

In classification-based stream mining, drift detection is essential in order to (i) inform operators when unintended system changes occur and (ii) make classifier updates more flexible when changes are intentional. Current detection approaches usually rely on the assumption that fully supervised labeled streams are available for monitoring (the changes in) classifier performance. This is an unrealistic scenario in many on-line real-world applications as true class labels would have to be known, which usually requires tedious feedback efforts of operators working with the systems. We propose two techniques to improve economy and applicability of current drift detection techniques: (i) a semi-supervised approach that employs single-pass active learning filters to select the most interesting samples for supervising classifier performance and (ii) a fully unsupervised approach based on the degree of overlap between a classifier’s output certainty distributions that can be applied to any unlabeled classification stream. For both variants, a specific handling of imbalanced class distributions in the streams is proposed, which allows also possible downtrends in classifier behavior for under-represented classes to be observed. The statistical monitoring of classifier behavior relies on a modified version of the Page-Hinkley test, where a fading factor and an automatic thresholding concept (based on the Hoeffding bound) were introduced to render it more flexible for detecting successive drift occurrences in a stream. We compared our approaches to the fully supervised variant in two real-world on-line applications, including a systematic analysis of the capabilities of our methods. The semi-supervised approach was able to detect real as well as artificially built-in drifts in these streams with a similar delay (of about 5–6 min) as the supervised variant, and this with only 20% actively selected samples. The unsupervised variant was able to detect input space drifts with reasonable delays as well, but failed to detect target concept drifts — using both approaches in tandem therefore allows us to distinguish between input space and target concept drifts.

A trust model for recommender agent systems

This paper aims to improve trust models in multi-agent systems based on four vital components, namely: reliability, similarity, satisfaction and trust transitivity. A number of different methods of computing these components were analyzed by considering the most representative existing trust models. The four trust components were identified from existing models then a trust model named trust transitivity---satisfaction---similarity---reliability (TtSSR) was proposed based on these components. TtSSR applied fuzzy logic for computing the identified components. Then by integrating the identified components and using Technique for Order of Preference by Similarity to Ideal Solution as a decision-making method, TtSSR selects the most trustworthy provider agent. The performance of TtSSR was compared with existing trust models using a simulator, specifically with Bayesian Network Model, Probability Certainty Distribution Model, and Dynamic Trust Model which are based on probability and TREPPS which is based on fuzzy logic. The experimental results revealed that TtSSR can significantly improve the accuracy of trust models; while the result of simulations demonstrated that the average accuracy of TtSSR in selecting a trustworthy agent is better than other models. Generally, the results indicated that when these four components were integrated, they performed significantly better in selecting a trustworthy agent as compared to other models.

Human-Water Relationship Evaluation Based on Cloud Model for Arid Area in Northwest China

Human-water relationship evaluation method based on cloud model is put forward to cope with the vagueness in evaluation language. With the intensive study of the specialty in human-water relationship in Northwest China arid area, three dimensions related to human-water relationship, natural water cycle’s health degree (NWHD), Social water cycle’s rationality degree (SWRD) and human-to-water fitness degree (HWFD) are selected to describe the harmony degree between water system and human system. The limit of evaluation index subordinate to each dimension corresponding to every harmony level is determined, and thus we can generate a comprehensive cloud model with evaluation index harmony level. When numeric values substitute for basic evaluation indexes, combined with the weights calculated from entropy weight method and AHP method, we get the certainty distributions of dimension indexes and integration objects among each harmony level. Case shows that human-water relationship evaluation method based on cloud model can effectively solve the vagueness and uncertainty in evaluation language. The evaluation results reflect the harmony degree in Northwest China arid area precisely.

On the rate allocation problem under co-existence of uncertain utility function parameters and uncertain link capacities

The allocation problem is considered for a computer communication network with network links being shared among multiple transmissions running at the same time. Typically, each transmission brings benefit to a network owner proportional to the transmission rate allowed. Since the link capacities are limited and the minimal rates defined by Quality of Service (QoS) requirements have to be preserved, the rate allocation problem is defined here as a problem of maximizing the overall benefit (utility) under capacity and QoS constraints. In this paper, we consider the uncertain version of the problem, in which unknown utility function parameters and unknown link capacities are treated as uncertain variables characterized by certainty distributions. We present a method for solving the problem in a unified way, with both uncertainties combined in a single certainty index.

A prior-free framework of coherent inference and its derivation of simple shrinkage estimators

The reasoning behind uses of confidence intervals and p-values in scientific practice may be made coherent by modeling the inferring statistician or scientist as an idealized intelligent agent. With other things equal, such an agent regards a hypothesis coinciding with a confidence interval of a higher confidence level as more certain than a hypothesis coinciding with a confidence interval of a lower confidence level. The agent uses different methods of confidence intervals conditional on what information is available. The coherence requirement means that all levels of certainty of hypotheses about the parameter agree with the same distribution of certainty over parameter space. The result is a unique and coherent fiducial distribution that encodes the post-data certainty levels of the agent.While many coherent fiducial distributions coincide with confidence distributions or Bayesian posterior distributions, there is a general class of coherent fiducial distributions that equates the two-sided p-value with the probability that the null hypothesis is true. The use of that class leads to point estimators and interval estimators that can be derived neither from the dominant frequentist theory nor from Bayesian theories that rule out data-dependent priors. These simple estimators shrink toward the parameter value of the null hypothesis without relying on asymptotics or on prior distributions.

Admission control in computer networks with uncertain parameters

PurposeThe purpose of this paper is to deal with a problem of admission control in computer networks when some of their parameters are uncertain. The case is considered when the most common probabilistic description of the uncertainty cannot be used and another approach should be applied.Design/methodology/approachThe uncertain versions of admission control problem with quality of service requirements are considered. The uncertain variables are used to describe possible values of the unknown parameters in computer networks.FindingsGiven are formulations for the admission control problem in computer networks with unknown values of the capacities based on the network utility maximization concept. Solution algorithms for all these problems are proposed.Research limitations/implicationsIt is assumed that an expert can describe possible values of uncertain network parameters in the form of a certainty distribution. Then the formalism of uncertain variables is applied and the knowledge of an expert is modelled with the use of certainty distributions. Decisions strongly depends on the quality of an expert's knowledge.Practical implicationsObtained admission control algorithms can be useful for planning and designing of computer networks.Originality/valueA new approach to the admission control problem in computer networks in the presence of uncertainty, in the case when the uncertain variable can be applied, is proposed and discussed.

QoS rate allocation in computer networks under uncertainty

Purpose – The purpose of this paper is to examine the problem of rate allocation (RA) in computer networks in cases when some parameters are unknown or their values are imprecise.Design/methodology/approach – The application of uncertain variables for the RA problems in computer networks in the presence of uncertainty is proposed.Findings – Decision‐making problem formulations for RA in computer networks with unknown parameters of the utility functions and bandwidths based on the network utility maximization concept are given. Solution algorithms for all these problems are proposed.Research limitations/implications – It is assumed that an expert can describe possible values of unknown network parameters in the form of a certainty distribution. Then, the formalism of uncertain variables is applied and the knowledge of an expert is modelled with certainty distributions.Practical implications – The RA algorithms obtained can be useful for designing and planning computer networks.Originality/value – The new a...

Seductive Simulations? Uncertainty Distribution Around Climate Models

This paper discusses the distribution of certainty around General Circulation Models (GCMs) – computer models used to project possible global climatic changes due to human emissions of greenhouse gases. It examines the trope of distance underpinning Donald MacKenzie’s concept of ‘certainty trough’, and calls for a more multi-dimensional and dynamic conceptualization of how uncertainty is distributed around technology. The certainty trough describes the level of certainty attached to particular technoscientific constructions as distance increases from the site of knowledge production, and proposes that producers of a given technology and its products are the best judges of their accuracy. Processes and dynamics associated with GCM modeling challenge the simplicity of the certainty trough diagram, mainly because of difficulties with distinguishing between knowledge producers and users, and because GCMs involve multiple sites of production. This case study also challenges the assumption that knowledge producers always are the best judges of the accuracy of their models. Drawing on participant observation and interviews with climate modelers and the atmospheric scientists with whom they interact, the study discusses how modelers, and to some extent knowledge producers in general, are sometimes less able than some users to identify shortcomings of their models.

Quality of cascade operations control based on uncertain variables

This article concerns a problem of control quality evaluation for a control plant composed of cascade operations. The execution time of each operation is described by an expert in the form of inequality containing an unknown parameter. All unknown parameters are assumed to be values of uncertain variables characterized by certainty distributions. This article gives a formulation and solution of the knowledge-based control problem, a proposition for a control quality index, a comprehensive example illustrating the approach presented, and the results of simulations.

APPLICATION OF UNCERTAIN VARIABLES TO STABILIZATION AND PARAMETRIC OPTIMIZATION OF UNCERTAIN DYNAMIC SYSTEMS

A class of dynamic discrete systems (control systems) with unknown parameters is considered. The unknown parameters are assumed to be values of uncertain variables described by an expert in the form of certainty distributions. The method of an estimation (evaluation) of the certainty index that the system is stable is presented and stabilization problems based on such an estimation are formulated. The analogous approach for the system with uncertain and random parameters is described. The method of a parametric optimization considered as a specific decision problem is proposed. Simple examples illustrate the presented approach.

Uncertain variables and their applications for control systems

The definitions and basic properties of so called uncertain variables are presented. The uncertain variables are described by certainty distributions given by an expert and characterizing approximate values of the variables. Control problems for uncertain systems with static and dynamic plants are considered. A method of the stability analysis for a system with uncertain parameters is described. Simple examples illustrate the presented approaches.

Application of Uncertain Variables to Control for a Class of Production Operations with Parametric Uncertainties

The paper concerns a class of production operations described by a relational knowledge representation with unknown parameters. The short description of so called uncertain variables is presented. The different formulations of the control problems are considered and the methods of the problems solving based on the certainty distributions given by an expert are described. The application of this approach to a raw material distribution among the parallel operations is proposed. Simple examples illustrate the presented approach.

Application of learning algorithms and uncertain variables in knowledge-based pattern recognition

A class of knowledge-based pattern recognition problems with relational and logical knowledge representation is formulated. In the first part, the learning algorithm consisting in using the learning sequence for astep-by-step evaluation, and updating of the knowledge is described. In the second part, the application of so-called uncertain variables and certainty distributions given by an expert is presented for an independent recognition problem and for recognition in a simple decision-making system. Simple examples and the results of simulations illustrating the general methods are included.

Uncertain variables and their applications for a class of uncertain systems

This paper presents definitions and basic properties of uncertain variables and shows how they may be applied to the analysis and decision making in static systems with unknown parameters in their mathematical models. The unknown parameters are assumed to be uncertain variables described by certainty distributions given by an expert. Two versions of the uncertain variables are described. The uncertain variable in the first version may be considered as a special case of the fuzzy number with a specific interpretation. Two forms of systems are considered in the paper: a system described by a function (functional system) and a system described by a relation (relational system). In both cases the statements and general solutions of the analysis and decision problems are presented. Simple examples illustrating the application of the uncertain variables are included.

Uncertain variables and their applications for a class of uncertain systems

This paper presents definitions and basic properties of uncertain variables and shows how they may be applied to the analysis and decision making in static systems with unknown parameters in their mathematical models. The unknown parameters are assumed to be uncertain variables described by certainty distributions given by an expert. Two versions of the uncertain variables are described. The uncertain variable in the first version may be considered as a special case of the fuzzy number with a specific interpretation. Two forms of systems are considered in the paper: a system described by a function (functional system) and a system described by a relation (relational system). In both cases the statements and general solutions of the analysis and decision problems are presented. Simple examples illustrating the application of the uncertain variables are included.