Neural Network-Based Uncertainty Quantification: A Survey of Methodologies and Applications

Kabir, H M Dipu; Khosravi, Abbas; Hosen, Mohammad Anwar; Nahavandi, Saeid

doi:10.1109/access.2018.2836917

Cited by 206 publications

(79 citation statements)

References 121 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In practical dynamic modeling, the inputs of control variables are usually point values, but the feedback variables of the previous outputs may be the intervals in the prediction application because the prediction values are intervals, which can reflect the predicting uncertainties to the future forecasting in the long-range prediction. That is not difficult for INNs but is hard to implement for the NN-based PIs mentioned in [16].…”

Section: Discussionmentioning

confidence: 99%

“…To address the problems that arise when modelling an uncertain dynamic plant based on UBB theory, we focus on a new interval modelling method by employing interval neural networks (INNs) in system design, which can effectively avoid problems such as the system models structural demands and the error compatibility requirement. Although several researchers have studied the development of prediction intervals (PIs) for NN forecasts [15], [16], the existing research on PIs is to use the point-valued neural network to realize the interval-valued prediction, which essentially belongs to statistical regression. Besides some limitations in understanding and difficulties in the training process of NN-based PI [16], it is very hard, or impossible to quantify the uncertainties in the system inputs and parameters, and it is difficult to provide dynamic long-range interval prediction, i.e., the previous outputs are fed back from the network itself to reflect the current point (interval) values to the future ones, while these are the necessary requirements of an uncertain dynamic process modeling.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Modeling Uncertain Dynamic Plants With Interval Neural Networks by Bounded-Error Data

2020

View full text Add to dashboard Cite

This paper presents a novel approach to building an interval dynamic model for an industrial plant with uncertainty by an interval neural network (INN). A new type of randomized learner model, named interval random vector functional-link network (IRVFLN), is proposed to take advantages of the inherent RVFLN in rapid modeling. The IRVFLN model is equipped with interval hidden input weights (and biases), which are randomly assigned from certain distribution/range and remain fixed, and the interval output weights can be evaluated by solving a couple of least squares problems. The comparative numerical experiments have verified the good potential of the proposed IRVFLN with the interval learner models produced by the error back-propagation algorithm. In the following modeling application, some measures for building IRVFLN with unknown but bounded (UBB) errors requirements are discussed in depth, in order to modeling an uncertain dynamic plant with IRVFLN by bounded-error data in either known or unknown error bounds. Finally, as a case study, the IRVFLN is applied to modeling a chemical interval dynamic plant with recycling, where the simulation results and generalization ability analysis demonstrate that the proposed method is suitable and effective. INDEX TERMS Interval neural network (INN), random vector functional-link network (RVFLN), unknown but bounded (UBB) errors, uncertain dynamic system modeling.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Modeling Uncertain Dynamic Plants With Interval Neural Networks by Bounded-Error Data

2020

View full text Add to dashboard Cite

show abstract

“…Ignorance can be described as a lack of knowledge, but the sociology of scientific ignorance refers to the ignorance of scientific research and the public ignorance of science [ 30 ]. Uncertainty refers to epistemic situations involving incomplete or unknown information to predict future events in many fields [ 31 ].…”

Section: Discourse Of a Systemic Resiliencementioning

confidence: 99%

A systemic approach to resilience and ecological sustainability during the COVID-19 pandemic: Human, societal, and ecological health as a system-wide emergent property in the Anthropocene

Zabaniotou

2020

Global Transitions

View full text Add to dashboard Cite

The COVID-19 pandemic has caused a worldwide disruption. Most people have never witnessed such a global threat, and the world’s leaders have not dealt with a crisis of this magnitude; moreover, Research & Innovation (R&I) teams have little time to invent new pharmaceutical therapies. Nations are trying to implement controlling strategies for avoiding significant losses, but this pandemic has already imprinted itself upon their citizens’ psyche, created social anxiety, and disrupted national economies. The complexity of the psychological, social, and economic interrelations of this new source of stress cannot be appropriately understood by scientific reductionism and specialised thinking only. It needs to be considered how the current pandemic links to questions of ecological sustainability and resilience. Further, we must rethink the complex interactions of human-nature health that drove the crisis, as proof of an unsustainable human civilisation. Accordingly, this paper aims to contribute to the transdisciplinary resilience dialogue on the health maintenance and life-supporting processes of the biosphere by focusing on the COVID-19 crisis. It explores various frameworks that are contributing to the transdisciplinary meta-perspective of resilience. Moreover, it proposes a humanistic approach based on not only controlling strategies involving containment and social isolation but also the ecological balance considering the human, societal, and ecological health as a system-wide emergent property. Conceptual frameworks of resilience are discussed—as mapping methodologies to structure the discourse—focusing on the role of leadership and empowerment. Furthermore, some positive insights are discussed, as a transdisciplinary integrator and solidarity facilitator of coping, mitigation, and decision-making in the time of uncertainty and anxiety created by the COVID-19 pandemic.

show abstract

“…Although computational intelligence methods exhibit adequate performance in estimation and prediction, uncertainty is not typically quantified by these modelling approaches, and only expected value is obtained. However, information on the dispersion of the output of the model provides more information about the phenomena modelled with uncertainty and more useful information from a decision-making point of view than the models with only expected value ( Kabir, Khosravi, Hosen, & Nahavandi, 2018;Shrivastava, Lohia, & Panigrahi, 2016 ).…”

Section: Introductionmentioning

confidence: 99%

Prediction interval methodology based on fuzzy numbers and its extension to fuzzy systems and neural networks

Marín

Cruz

Sáez

et al. 2019

Expert Systems with Applications

View full text Add to dashboard Cite

Prediction interval modelling has been proposed in the literature to characterize uncertain phenomena and provide useful information from a decision-making point of view. In most of the reported studies, assumptions about the data distribution are made and/or the models are trained at one step ahead, which can decrease the quality of the interval in terms of the information about the uncertainty modelled for a higher prediction horizon. In this paper, a new prediction interval modelling methodology based on fuzzy numbers is proposed to solve the abovementioned drawbacks. Fuzzy and neural network prediction interval models are developed based on this proposed methodology by minimizing a novel criterion that includes the coverage probability and normalized average width. The fuzzy number concept is considered because the affine combination of fuzzy numbers generates, by definition, prediction intervals that can handle uncertainty without requiring assumptions about the data distribution. The developed models are compared with a covariance-based prediction interval method, and high-quality intervals are obtained, as determined by the narrower interval width of the proposed method. Additionally, the proposed prediction intervals are tested by forecasting up to two days ahead of the load of the Huatacondo microgrid in the north of Chile and the consumption of the residential dwellings in the town of Loughborough, UK. The results show that the proposed models are suitable alternatives to electrical consumption forecasting because they obtain the minimum interval widths that characterize the uncertainty of this type of stochastic process. Furthermore, the information provided by the obtained prediction interval could be used to develop robust energy management systems that, for example, consider the worst-case scenario.

show abstract

Neural Network-Based Uncertainty Quantification: A Survey of Methodologies and Applications

Cited by 206 publications

References 121 publications

Modeling Uncertain Dynamic Plants With Interval Neural Networks by Bounded-Error Data

Modeling Uncertain Dynamic Plants With Interval Neural Networks by Bounded-Error Data

A systemic approach to resilience and ecological sustainability during the COVID-19 pandemic: Human, societal, and ecological health as a system-wide emergent property in the Anthropocene

Prediction interval methodology based on fuzzy numbers and its extension to fuzzy systems and neural networks

Contact Info

Product

Resources

About