Probabilistic Modeling Framework Research Articles

Towards resilience of offshore wind farms: A framework and application to asset integrity management

This paper develops a general probabilistic framework for resilience modeling and analysis of offshore wind farm (OWF), and illustrates how such a framework may be implemented within the modeling techniques and tools commonly applied in the industry. Based on this framework the significance of prevailing uncertainties and the effects of different decision alternatives relevant in the context of asset integrity management (AIM) are studied and discussed. In the framework, OWFs are modeled as system-of-systems by a hierarchical model where the life-cycle performances of each system, as well as the dependencies between these systems, are represented probabilistically. The quantification of resilience is undertaken based on a scenario-based modeling of life cycle benefits and costs in which resilience failure is defined as the exhaustion of the economic capacity accumulated by the system over time. Moreover, this paper introduces resilience-informed decision-making for OWF in the context of AIM. The proposed framework is applied to the OWFs populated with NREL 5MW offshore wind turbines (OWTs). Events of typhoon-induced waves and winds are considered as the two random environmental load processes affecting the OWF’s dynamic responses and for which their resilience performances are carried out. Finally, the resilience performances of the OWFs are studied and discussed for a range of decision alternatives relevant to AIM.

Bayesian Methods for Inferring Missing Data in the BATSE Catalog of Short Gamma-Ray Bursts

The knowledge of the redshifts of Short-duration Gamma-Ray Bursts (SGRBs) is essential for constraining their cosmic rates and thereby the rates of related astrophysical phenomena, particularly Gravitational Wave Radiation (GWR) events. Many of the events detected by gamma-ray observatories (e.g., BATSE, Fermi, and Swift) lack experimentally measured redshifts. To remedy this, we present and discuss a generic data-driven probabilistic modeling framework to infer the unknown redshifts of SGRBs in the BATSE catalog. We further explain how the proposed probabilistic modeling technique can be applied to newer catalogs of SGRBs and other astronomical surveys to infer the missing data in the catalogs.

버티컬 검색에서의 확률 그래픽 모델 기반 클릭모형

Modern search engines provide the result including not only the traditional web documents, but also the segments aggregated with specialized information such as images, videos, and news, called vertical, to the users. Most click models associated with vertical search were designed to reduce various biases caused by verticals in search engine result pages to evaluate user behavior of web documents. However, evaluation of verticals themselves has not been conducted. In this paper, we propose vertical click models to analyze user behavior associated with both verticals and documents. The proposed click models based on probabilistic graphical model framework directly reflect events occurred in verticals and documents, so that they allow to elaborately calculate and evaluate the preference of verticals felt by the user. The proposed models were evaluated with a click log dataset from the popular commercial search engine in Korea. The results demonstrated superiority of our model over existing models in terms of normalized discounted cumulative gain.

Probabilistic framework for the parametric modeling of leakages in water distribution networks: large scale application to the City of Patras in Western Greece

Probabilistic framework for the parametric modeling of leakages in water distribution networks: large scale application to the City of Patras in Western Greece

Jointly Modeling Rating Responses and Times with Fuzzy Numbers: An Application to Psychometric Data

In several research areas, ratings data and response times have been successfully used to unfold the stagewise process through which human raters provide their responses to questionnaires and social surveys. A limitation of the standard approach to analyze this type of data is that it requires the use of independent statistical models. Although this provides an effective way to simplify the data analysis, it could potentially involve difficulties with regard to statistical inference and interpretation. In this sense, a joint analysis could be more effective. In this research article, we describe a way to jointly analyze ratings and response times by means of fuzzy numbers. A probabilistic tree model framework has been adopted to fuzzify ratings data and four-parameters triangular fuzzy numbers have been used in order to integrate crisp responses and times. Finally, a real case study on psychometric data is discussed in order to illustrate the proposed methodology. Overall, we provide initial findings to the problem of using fuzzy numbers as abstract models for representing ratings data with additional information (i.e., response times). The results indicate that using fuzzy numbers leads to theoretically sound and more parsimonious data analysis methods, which limit some statistical issues that may occur with standard data analysis procedures.

A two-sample robust Bayesian Mendelian Randomization method accounting for linkage disequilibrium and idiosyncratic pleiotropy with applications to the COVID-19 outcomes.

Mendelian randomization (MR) is a statistical method exploiting genetic variants as instrumental variables to estimate the causal effect of modifiable risk factors on an outcome of interest. Despite wide uses of various popular two-sample MR methods based on genome-wide association study summary level data, however, those methods could suffer from potential power loss or/and biased inference when the chosen genetic variants are in linkage disequilibrium (LD), and also have relatively large direct effects on the outcome whose distribution might be heavy-tailed which is commonly referred to as the idiosyncratic pleiotropy phenomenon. To resolve those two issues, we propose a novel Robust Bayesian Mendelian Randomization (RBMR) model that uses the more robust multivariate generalized -distribution to model such direct effects in a probabilistic model framework which can also incorporate the LD structure explicitly. The generalized -distribution can be represented as a Gaussian scaled mixture so that our model parameters can be estimated by the expectation maximization (EM)-type algorithms. We compute the standard errors by calibrating the evidence lower bound using the likelihood ratio test. Through extensive simulation studies, we show that our RBMR has robust performance compared with other competing methods. We further apply our RBMR method to two benchmark data sets and find that RBMR has smaller bias and standard errors. Using our proposed RBMR method, we find that coronary artery disease is associated with increased risk of critically ill coronavirus disease 2019. We also develop a user-friendly R package RBMR (https://github.com/AnqiWang2021/RBMR) for public use.

Evolving probabilistic modeling for long-term significant wave heights with a focus on extremes

The present paper aims to formulate a novel framework of probabilistic model for long-term significant wave height to facilitate accurate probabilistic analysis and further decision making in engineering practice. It is found that there are several sources of inconsistency through the investigations of the records of significant wave height and correspondingly, three main challenges in the probabilistic modeling together with the relevant characteristics of long-term significant wave height are presented. A cluster based probabilistic model, in which the extremes are highlighted, is derived afterwards by optimization to ensure that it keeps consistent with the information conveyed from the records. Further, the model will also evolve with the recognition of the inconsistency between the records collected from different lengths of period. Finally, a framework for the analysis is formulated and its application is illustrated.

Modeling a class of thermal ice probes for accessing the solar system’s ocean worlds

The subsurface oceans of icy moons in the outer solar system are prime candidates for harboring extra-terrestrial life. Missions to melt through the ice shells of these worlds utilizing an ice probe or “cryobot” to access the oceans have been proposed. To design such a cryobot for a successful mission, the relationship between a cryobot’s heat budget and descent speed within any given environment must be understood. Large uncertainties in the ice shell thickness and temperature profile require that the problem be treated via probabilistic techniques, such as a Monte Carlo simulation. By contrast, the speed of a cryobot descending through ice with specified far-field temperature and material properties can, in principle, be quantified with little uncertainty. A model of cryobot performance that can be computed quickly and accurately will save significant computational time for the broader simulation, and allow for the mission’s basic thermal needs to be understood. The work reported here builds upon an existing analytic thermal model of ice probes that, while suitable for terrestrial conditions, begins to incur potentially problematic errors in extraterrestrial environments. The new model provides improvement to the previous model for flight-relevant probes in extra-terrestrial ice conditions, predicting the total heat budget to within 1.2% of high-fidelity numerical simulations while requiring a small fraction of the computational time. The model can be non-dimensionalized in a compact form, making it suitable for modeling a broad class of cylindrical ice probes in a probabilistic modeling framework for determining the time required to access the sub-surface ocean. Such models can then provide the high-level requirements for cryobot system architectures that will enable successful ocean-access missions.

An Integrated, Probabilistic Modeling Approach to Assess the Evolution of Barrier-Island Systems Over the Twenty-First Century

Barrier-island systems, spanning ∼7% of the world’s coastlines, are of great importance to society because not only they provide attractive, liveable space for coastal communities but also act as the first line of defense from coastal storms. As many of these unique coastal systems are affected by both oceanic and terrestrial processes, it is necessary to consider the holistic behavior of applicable terrestrial and coastal processes when investigating their evolution under plausible future scenarios for climate change, population growth and human activities. Such holistic assessments, also accounting for uncertainties, can readily be achieved via reduced-complexity modeling techniques, owing to their much faster simulation times compared to sophisticated process-based models. Here, we develop and demonstrate a novel probabilistic modeling framework to obtain stochastic projections of barrier-island evolution over the twenty-first century while accounting for relevant oceanic and terrestrial processes under climate change impacts and anthropogenic activities. The model is here demonstrated at the Chandeleur islands (Louisiana, United States) under the four Intergovernmental Panel on Climate Change (IPCC) greenhouse gas emission scenarios (i.e., Representative Concentration Pathways 2.6, 4.5, 6.0, and 8.5) with results indicating that there are significant uncertainties in projected end-century barrier-island migration distance and available barrier freeboard under the high emission scenario RCP 8.5. The range of uncertainties in these projections underscores the value of stochastic projections in supporting the development of effective adaptation strategies for these fragile coastal systems.

ACE: a probabilistic model for characterizing gene-level essentiality in CRISPR screens

High-throughput CRISPR-Cas9 knockout screens are widely used to evaluate gene essentiality in cancer research. Here we introduce a probabilistic modeling framework, Analysis of CRISPR-based Essentiality (ACE), that accounts for multiple sources of variation in CRISPR-Cas9 screens and enables new statistical tests for essentiality. We show using simulations that ACE is effective at predicting both absolute and differential essentiality. When applied to publicly available data, ACE identifies known and novel candidates for genotype-specific essentiality, including RNA m6-A methyltransferases that exhibit enhanced essentiality in the presence of inactivating TP53 mutations. ACE provides a robust framework for identifying genes responsive to subtype-specific therapeutic targeting.

Overcoming the Cold Start Problem of Customer Relationship Management Using a Probabilistic Machine Learning Approach

The success of customer relationship management programs ultimately depends on the firm's ability to identify and leverage differences across customers—a difficult task when firms attempt to manage new customers, for whom only the first purchase has been observed. The lack of repeated observations for these customers poses a structural challenge for firms to infer unobserved differences across them. This is what the authors call the “cold start” problem of customer relationship management, whereby companies have difficulties leveraging existing data when they attempt to make inferences about customers at the beginning of their relationship. The authors propose a solution to the cold start problem by developing a probabilistic machine learning modeling framework that leverages the information collected at the moment of acquisition. The main aspect of the model is that it flexibly captures latent dimensions that govern the behaviors observed at acquisition as well as future propensities to buy and to respond to marketing actions using deep exponential families. The model can be integrated with a variety of demand specifications and is flexible enough to capture a wide range of heterogeneity structures. The authors validate their approach in a retail context and empirically demonstrate the model's ability to identify high-value customers as well as those most sensitive to marketing actions right after their first purchase.

Nature-based solutions for flood risk reduction: A probabilistic modeling framework

<h2>Summary</h2> Flooding is the most frequent and damaging natural hazard globally. While nature-based solutions can reduce flood risk, they are not part of mainstream risk management. We develop a probabilistic risk analysis framework to quantify these benefits that (1) accounts for frequent small events and rarer large events, (2) can be applied to large basins and data-scarce contexts, and (3) quantifies economic benefits and reduction in people affected. Measuring benefits in terms of avoided losses enables the integration of nature-based solutions in standard cost-benefit analysis of protective infrastructure. Results for the Chindwin River basin in Myanmar highlight the potential consequences of deforestation on long-term flood risk. We find that loss reduction is driven by small but frequent storms, suggesting that current practice relying on large storms may underestimate the benefits of nature-based solutions. By providing average annual losses, the framework helps mainstream nature-based solutions in infrastructure planning or insurance practice.

A probabilistic and anisotropic failure metric for ascending thoracic aortic aneurysm risk assessment

To noninvasively assess the risk of aneurysm rupture and dissection, an accurate material failure metric of the aortic wall is crucial. Previous studies used deterministic or isotropic failure metrics for the aortic wall. However, experimental studies have shown that aortic wall tensile strengths in circumferential and axial directions are significantly different (i.e., anisotropic) and vary greatly among patients. In this study, we developed a new probabilistic and anisotropic material failure metric for rupture risk assessment of ascending thoracic aortic aneurysm (ATAA). We performed uniaxial tensile failure tests using aortic tissue samples of 84 ATAA patients, from which a joint probability distribution of the anisotropic wall strengths was obtained. Subsequently, we derived an anisotropic failure probability (FP) metric based on the Tsai-Hill (TH) failure criterion. The novel FP metric incorporates uncertainty and anisotropy of failure properties. To compare the FP metric with traditional deterministic and isotropic metrics, we numerically estimated “baseline” risks of additional 41 ATAA patients using matching CT images and tissue testing data. We presented different risk assessment methods (e.g., with and without patient-specific hyperelastic properties) and compared them using receiver operating characteristic (ROC) curves. The results demonstrated that: (1) the probabilistic FP metric outperforms the deterministic TH metric and the isotropic maximum principal stress; (2) patient-specific hyperelastic properties can help to improve the performance of probabilistic FP metric in ATAA risk assessment. The proposed probabilistic modeling framework may be adopted for other types of materials.

COVID-19 non-pharmaceutical intervention portfolio effectiveness and risk communication predominance

Non-pharmaceutical interventions (NPIs) including resource allocation, risk communication, social distancing and travel restriction, are mainstream actions to control the spreading of Coronavirus disease 2019 (COVID-19) worldwide. Different countries implemented their own combinations of NPIs to prevent local epidemics and healthcare system overloaded. Portfolios, as temporal sets of NPIs have various systemic impacts on preventing cases in populations. Here, we developed a probabilistic modeling framework to evaluate the effectiveness of NPI portfolios at the macroscale. We employed a deconvolution method to back-calculate incidence of infections and estimate the effective reproduction number by using the package EpiEstim. We then evaluated the effectiveness of NPIs using ratios of the reproduction numbers and considered them individually and as a portfolio systemically. Based on estimates from Japan, we estimated time delays of symptomatic-to-confirmation and infection-to-confirmation as 7.4 and 11.4 days, respectively. These were used to correct surveillance data of other countries. Considering 50 countries, risk communication and returning to normal life were the most and least effective yielding the aggregated effectiveness of 0.11 and − 0.05 that correspond to a 22.4% and 12.2% reduction and increase in case growth. The latter is quantified by the change in reproduction number before and after intervention implementation. Countries with the optimal NPI portfolio are along an empirical Pareto frontier where mean and variance of effectiveness are maximized and minimized independently of incidence levels. Results indicate that implemented interventions, regardless of NPI portfolios, had distinct incidence reductions and a clear timing effect on infection dynamics measured by sequences of reproduction numbers. Overall, the successful suppression of the epidemic cannot work without the non-linear effect of NPI portfolios whose effectiveness optimality may relate to country-specific socio-environmental factors.

Variational Bayesian probabilistic modeling framework for data-driven distributed process monitoring

Data-driven process monitoring has gained increasing attention because of the increasing demand in process safety and the rapid advancement of data gathering techniques. When monitoring a plant-wide multiunit process, establishing a monitor for each unit individually ignores the correlations among units, whereas establishing a global monitor for the entire process ignores the local process behavior. A variational Bayesian-based probabilistic modeling approach is proposed for efficient distributed process monitoring. A novel probabilistic latent variable model is developed to characterize the variable relationship in each local unit and among units. First, variational Bayesian-based latent variable extraction is performed in each local unit, through which variable relationship within a local unit is characterized. Second, variational Bayesian-based regression model is established between the latent variables and neighboring variables, through which the variable relationship among units is characterized. Then, modeling residuals and monitoring statistics are generated, through which the process status and the type of a detected fault are identified. The effectiveness of the proposed probabilistic modeling and monitoring method is verified by three case studies, including a numerical example, the Tennessee Eastman benchmark process, and a laboratory distillation process.

Probabilistic identification of rockfall source areas at regional scale in El Hierro (Canary Islands, Spain)

Modelling rockfall phenomena is complex and requires various inputs, including an accurate location of the source areas. Source areas are controlled by geomorphological, geological, or other geo-environmental factors and may largely influence the results of the modelling. In the Canary Islands, rockfalls are extremely common and pose a major threat to society, costing lives, disrupting infrastructure, and destroying livelihoods. In 2011, the volcanic event on the island of El Hierro triggered numerous rockfalls that affected strategic infrastructures, with a substantial impact on the local population. During the emergency, the efforts performed to map the source areas and to model the rockfalls in the considerably steep landscape characterising the island were not trivial. To better identify the rockfall source areas, we propose a probabilistic modelling framework that applies a combination of multiple statistical models using the source area locations mapped in the field as the dependent variable and a set of thematic data as independent variables. The models use as input morphometric parameters derived from the Digital Elevation Model and lithological data as an expression of the mechanical behaviour of the rocks. The analysis of different training and validation scenarios allowed us to test the model sensitivity to the input data, select the optimal model training configuration, and evaluate the model applicability outside the training areas. The final map obtained from the model for the entire island of El Hierro provides the probability of a given location being a potential source area and can be used as the input for rockfall runout simulation modelling.

Probabilistic Models with Deep Neural Networks.

Recent advances in statistical inference have significantly expanded the toolbox of probabilistic modeling. Historically, probabilistic modeling has been constrained to very restricted model classes, where exact or approximate probabilistic inference is feasible. However, developments in variational inference, a general form of approximate probabilistic inference that originated in statistical physics, have enabled probabilistic modeling to overcome these limitations: (i) Approximate probabilistic inference is now possible over a broad class of probabilistic models containing a large number of parameters, and (ii) scalable inference methods based on stochastic gradient descent and distributed computing engines allow probabilistic modeling to be applied to massive data sets. One important practical consequence of these advances is the possibility to include deep neural networks within probabilistic models, thereby capturing complex non-linear stochastic relationships between the random variables. These advances, in conjunction with the release of novel probabilistic modeling toolboxes, have greatly expanded the scope of applications of probabilistic models, and allowed the models to take advantage of the recent strides made by the deep learning community. In this paper, we provide an overview of the main concepts, methods, and tools needed to use deep neural networks within a probabilistic modeling framework.

Stochastic hydrology: the development of main ideas in Russia

The aim of the work is to assess the current state of scientific research and the main results in the field of stochastic hydrology. Various problems of scientific and applied hydrology are solved on the basis of different types of stochastic models. On the basis of these models, methods are being developed for obtaining secure engineering runoff characteristics, or reliable hydrological forecasts of various lead times. Most of the water management, construction, environmental and other problems are solved today within the framework of probabilistic models and statistical methods. The main tasks of stochastic hydrology can be divided into several groups. Despite the discussion in hydrology on the applicability of the theory of probability and random processes to the analysis of fluctuations of hydrometeorological series, the problem of choosing a research methodology remains one of the most important, and is often discussed in connection with anomalous manifestations of hydrological phenomena (extreme floods, prolonged droughts). An important applied and theoretical problem in hydrology is the choice of a one-dimensional probability distribution. In this part, discussions are important both on the problem of the type of new distributions and on methods for estimating parameters. Ideas and hypotheses are considered that can significantly increase the reliability of statistical characteristics. Climatic changes in many cases cause disturbances in the stationarity of time series of hydrological characteristics. Progress in the development of new approaches to probabilistic modeling is associated both with the well-known problem of the sufficiency of the Markov hypothesis and with the need to introduce new, more complex models and Bayesian estimation methods. The article is devoted to the discussion of the current state of these problems in the form of a scientific review.

Probabilistic Resistive Switching Device Modeling Based on Markov Jump Processes

In this work, a versatile mathematical framework for multi-state probabilistic modeling of Resistive Switching (RS) devices is proposed for the first time. The mathematical formulation of memristor and Markov jump processes are combined and, by using the notion of master equations for finite-states, the inherent probabilistic time-evolution of RS devices is sufficiently modeled. In particular, the methodology is generic enough and can be applied for $N$ states; as a proof of concept, the proposed framework is further stressed for both two-state RS paradigm, namely $N=2$, and multi-state devices, namely $N=4$. The presented I-V results demonstrate in a qualitative and quantitative manner, adequate matching with other modeling approaches.

Improvement Support Vector Machine Using Genetic Algorithm in Farmers Term of Trade Prediction at Central Java Indonesia

The welfare of farmers is a strategic problem in Indonesia. The Farmer term of the trade (FTT) is one indicator to measure the welfare of farmers. FTT is a measurement of the comparison of the price index received by farmers (It) with the price index paid by farmers (Ib). Some models for FTT prediction on previous research are using ANN, SVM, MLR, Markov Chain - Predictive Probabilistic Architecture Modeling Framework (P2AMF), Singular Spectrum Analysis (SSA) - ARIMA and ANN-PSO. Previous FFT research in 2018 used three prediction methods, namely using the ANN, SVM and MLR algorithms with the best RMSE being 0.00098. Then in 2019, FTT research was followed by optimization of the ANN parameters using PSO for weighting (ANN-PSO) and obtaining the best RMSE was 0.00062. This study evaluates the robustness of the prediction models of FTTs in the Central Java region using SVM. Then proceed with increasing SVM prediction accuracy using GA (SVM-GA). SVM-GA has resulted in an increase in FTT prediction accuracy. This study has found that the SVM method has better robustness than the ANN method. The development of research related to the accuracy for the FTT prediction model in the Central Java Province has increased, starting from 2018 with RMSE 0.00098; in 2019 with RMSE 0.00062 and the results of this study resulted in the best RMSE of 0.00037.