Probabilistic Modeling Approach Research Articles

Stochastic Modeling of Temporal Enhanced Ultrasound: Impact of Temporal Properties on Prostate Cancer Characterization.

Temporal enhanced ultrasound (TeUS) is a new ultrasound-based imaging technique that provides tissue-specific information. Recent studies have shown the potential of TeUS for improving tissue characterization in prostate cancer diagnosis. We study the temporal properties of TeUS-temporal order and length-and present a new framework to assess their impact on tissue information. We utilize a probabilistic modeling approach using hidden Markov models (HMMs) to capture the temporal signatures of malignant and benign tissues from TeUS signals of nine patients. We model signals of benign and malignant tissues (284 and 286 signals, respectively) in their original temporal order as well as under order permutations. We then compare the resulting models using the Kullback-Liebler divergence and assess their performance differences in characterization. Moreover, we train HMMs using TeUS signals of different durations and compare their model performance when differentiating tissue types. Our findings demonstrate that models of order-preserved signals perform statistically significantly better (85% accuracy) in tissue characterization compared to models of order-altered signals (62% accuracy). The performance degrades as more changes in signal order are introduced. Additionally, models trained on shorter sequences perform as accurately as models of longer sequences. The work presented here strongly indicates that temporal order has substantial impact on TeUS performance; thus, it plays a significant role in conveying tissue-specific information. Furthermore, shorter TeUS signals can relay sufficient information to accurately distinguish between tissue types. Understanding the impact of TeUS properties facilitates the process of its adopting in diagnostic procedures and provides insights on improving its acquisition.

Quantitative analysis of breast cancer diagnosis using a probabilistic modelling approach

BackgroundBreast cancer is the most prevalent cancer in women in most countries of the world. Many computer-aided diagnostic methods have been proposed, but there are few studies on quantitative discovery of probabilistic dependencies among breast cancer data features and identification of the contribution of each feature to breast cancer diagnosis. MethodsThis study aims to fill this void by utilizing a Bayesian network (BN) modelling approach. A K2 learning algorithm and statistical computation methods are used to construct BN structure and assess the obtained BN model. The data used in this study were collected from a clinical ultrasound dataset derived from a Chinese local hospital and a fine-needle aspiration cytology (FNAC) dataset from UCI machine learning repository. ResultsOur study suggested that, in terms of ultrasound data, cell shape is the most significant feature for breast cancer diagnosis, and the resistance index presents a strong probabilistic dependency on blood signals. With respect to FNAC data, bare nuclei are the most important discriminating feature of malignant and benign breast tumours, and uniformity of both cell size and cell shape are tightly interdependent. ContributionsThe BN modelling approach can support clinicians in making diagnostic decisions based on the significant features identified by the model, especially when some other features are missing for specific patients. The approach is also applicable to other healthcare data analytics and data modelling for disease diagnosis.

Modeling and Quantification of Model-Form Uncertainties in Eigenvalue Computations Using a Stochastic Reduced Model

A feasible, nonparametric, probabilistic approach for modeling and quantifying model-form uncertainties associated with a computational model designed for the solution of a generalized eigenvalue problem is presented. It is based on the construction of a stochastic, projection-based reduced-order model associated with a high-dimensional model using three innovative ideas: 1) the substitution of the deterministic reduced-order basis with a stochastic counterpart featuring a reduced number of hyperparameters, 2) the construction of this stochastic reduced-order basis on a subset of a compact Stiefel manifold to guarantee the linear independence of its column vectors and the satisfaction of any constraints of interest, and 3) the formulation and solution of a reduced-order inverse statistical problem to determine the hyperparameters so that the mean value and statistical fluctuations of the eigenvalues predicted using the stochastic, projection-based reduced-order model match target values obtained from available data. Consequently, the proposed approach for modeling model-form uncertainties can be interpreted as an effective approach for extracting from data fundamental information and/or knowledge that are not captured by a deterministic computational model, and incorporating them in this model. Its potential for quantifying model-form uncertainties in generalized eigencomputations is demonstrated for a natural vibration analysis of a small-scale replica of an X-56-type aircraft made of a composite material for which ground-vibration-test data are available.

Exploration of the Bayesian Network framework for modelling window control behaviour

Extended literature reviews confirm that the accurate evaluation of energy-related occupant behaviour is a key factor for bridging the gap between predicted and actual energy performance of buildings. One of the key energy-related human behaviours is window control behaviour that has been modelled by different probabilistic modelling approaches. In recent years, Bayesian Networks (BNs) have become a popular representation based on graphical models for modelling stochastic processes with consideration of uncertainty in various fields, from computational biology to complex engineering problems. This study investigates the potential applicability of BNs to capture underlying complicated relationships between various influencing factors and energy-related behavioural actions of occupants in buildings: in particular, window opening/closing behaviour of occupants in residential buildings is investigated. This study addresses five key research questions related to modelling window control behaviour: (A) variable selection for identifying key drivers impacting window control behaviour, (B) correlations between key variables for structuring a statistical model, (C) target definition for finding the most suitable target variable, (D) BN model with capabilities to treat mixed data, and (E) validation of a stochastic BN model. A case study on the basis of measured data in one residential apartment located in Copenhagen, Denmark provides key findings associated with the five research questions through the modelling process of developing the BN model.

An investigation of the selection strategies impact on MOEDAs: CMA-ES and UMDA

Researchers have acknowledged the importance of the selection, replacement, and archiving strategies in the behavior of evolutionary algorithms (EAs). However, these important relationships have not been deeply investigated in terms of Estimation of Distribution algorithms (EDAs). In a preliminary research, we focused on UMDA, a simple EDA that uses univariate Gaussian factorizations. Experimental results confirm that the choice of the selection method can provoke probabilistic modeling to be more effective for some classes of functions. Motivated by these results, this study is extended by evaluating several variants of selection strategies and probabilistic modeling approaches. The aim is to detect possible interactions between these two important components of evolutionary algorithms. Specifically, we use the selection strategies as defined for NSGA2, SPEA2, and IBEA algorithms, and the probabilistic models implemented as part of UMDA and CMA-ES, and a simple crossover operator (SBX). The recently introduced COCO framework comprising 55 bi-objective functions is used as the benchmark for the analysis. The results show that probabilistic modeling has an advantage over the classical genetic operator regardless of the selection method applied. Nevertheless, the results also show that some selection methods have a better performance when applied together with EDAs.

Probabilistic Modeling Approach to Reducing Healthcare Costs With Reflex Testing.

Statistical methods can be utilized to optimize the order for reflex diagnostic testing to attenuate patient and hospital costs without affecting quality of care. Our objective is to demonstrate the method of developing an order for testing and to apply this method to an illustrative example. An algorithm was developed for minimizing costs for any given number of diagnostic tests, and it was retrospectively applied to a sample. The actual scenario of using both tests on all patients was compared to 2 other hypothetical reflex testing approaches: all patients are given 1 test, and those patients who tested negative were then given the second test. The 2 scenarios would have saved 37.1% and 17.4% in testing costs, respectively. These calculations could be applied to numerous situations to reduce costs for patients and hospitals. We propose that this methodology would be best used in conjunction with any existing quality improvement initiatives.

Effects of lecture attendance, aptitude, individual heterogeneity and pedagogic intervention on student performance: A probability model approach

This article uses a logistic probability distribution approach to examine the effect of lecture attendance, aptitude test results, individual heterogeneity and pedagogic intervention on student performance for first-year microeconomics and second-year macroeconomics modules at a leading South African university. The research was motivated by the throughput concerns in South African institutions of higher education, where approximately one in four of the students enrolled complete their degrees in the minimum regulated time. Using secondary data of 630 and 360 first- and second-year students, respectively, the findings revealed that lecture attendance, aptitude score and having received a foreign high school education have a positive and statistically significant effect on academic performance. Students who received intervention and those using English at home performed better than others at second-year level. Based on these findings, a more stringent attendance policy, varied assessment measures, and wider intervention strategies, among others, are recommended to enhance performance.

The analytical limits of modeling short diffusion timescales

Diffusion modeling of chemical zonation in crystals and glasses is a widely used method of quantifying the timescales of a variety of geological processes. Obtaining timescale information from diffusion modeling relies on fitting modeled diffusion profiles to measured compositional gradients. Thus analytical factors, such as spatial resolution, analysis location and analytical uncertainty have the potential to limit the accuracy and precision of calculated diffusion timescales, especially when the resolution of the individual analyses approaches the width of the observed diffusion gradient. Herein we use a probabilistic modeling approach to assess the accuracy of timescales based on diffusion modeling at various spatial resolutions and diffusivities. We have calculated synthetic 1D diffusion profiles produced from simple step function geometries at various diffusivities and timescales and then resampled these at common analytical spatial resolutions. Standard diffusion modeling techniques were used to estimate apparent diffusion timescales from the resampled profiles to compare with the “true” synthetic times. Results confirm that for a given diffusivity, higher analytical spatial resolution gives access to shorter timescales, and that for each analytical resolution there is a minimum threshold timescale, below which diffusion modeling significantly overestimates the true timescale. The accuracy of short diffusion timescales depends on the width of the diffusion gradient (the portion of a profile that has been modified by diffusion) relative to the analytical spatial resolution, with models becoming accurate (within 20%) at a gradient width/spot size ratio>2. The precision of diffusion timescales depends on the gradient width to spot size ratio, the magnitude of analytical uncertainty and the magnitude of the compositional difference across the step function boundary. Precision is improved with a large gradient width/spot size ratio, and where the magnitude of the compositional difference across the step function boundary is large relative to the analytical uncertainty. We present a generalized method to quantify these threshold timescales for any given spatial resolution and diffusivity, and present simple guidelines to aid in selecting compositional profiles that will produce more accurate diffusion timescale estimates.

Metabolic network segmentation: A probabilistic graphical modeling approach to identify the sites and sequential order of metabolic regulation from non-targeted metabolomics data.

In recent years, the number of large-scale metabolomics studies on various cellular processes in different organisms has increased drastically. However, it remains a major challenge to perform a systematic identification of mechanistic regulatory events that mediate the observed changes in metabolite levels, due to complex interdependencies within metabolic networks. We present the metabolic network segmentation (MNS) algorithm, a probabilistic graphical modeling approach that enables genome-scale, automated prediction of regulated metabolic reactions from differential or serial metabolomics data. The algorithm sections the metabolic network into modules of metabolites with consistent changes. Metabolic reactions that connect different modules are the most likely sites of metabolic regulation. In contrast to most state-of-the-art methods, the MNS algorithm is independent of arbitrary pathway definitions, and its probabilistic nature facilitates assessments of noisy and incomplete measurements. With serial (i.e., time-resolved) data, the MNS algorithm also indicates the sequential order of metabolic regulation. We demonstrated the power and flexibility of the MNS algorithm with three, realistic case studies with bacterial and human cells. Thus, this approach enables the identification of mechanistic regulatory events from large-scale metabolomics data, and contributes to the understanding of metabolic processes and their interplay with cellular signaling and regulation processes.

Sensitivity analysis of parametric uncertainties and modeling errors in computational-mechanics models by using a generalized probabilistic modeling approach

Engineering analyses of structures may be confronted with many sources of uncertainty, which may be of different types, such as parametric uncertainties versus modeling errors, and which may pertain to different structural components when complex structures are analyzed. Soize, Generalized probabilistic approach of uncertainties in computational dynamics using random matrices and polynomial chaos decompositions, Int. J. Numer. Meth. Eng., 81:939–970, 2010 has recently introduced a generalized probabilistic modeling approach, which can individually represent parametric uncertainties and modeling errors and which can individually represent sources of uncertainty pertaining to different structural components of complex structures. In this paper, we propose to augment this generalized probabilistic modeling approach with a stochastic sensitivity analysis in order to quantify and gain insight into separate impacts of distinct sources of uncertainty on quantities of interest. We demonstrate the proposed methodology by applying it to two computational-mechanics models involving uncertainty.

Understanding Land Use, Land Cover and Woodland-Based Ecosystem Services Change, Mabalane, Mozambique

Charcoal production constitutes a key ecosystem service in Mozambique, with an estimated market value of US$400 million a year. Due to the central role the charcoal industry plays in local livelihoods, availability of suitable wood for charcoal production has decreased because of changes in land use and land cover (LULC). This paper applied a probabilistic modelling approach combining Bayesian Belief Networks (BBNs), Geographic Information Systems, Remote Sensing data, field data, and expertise from different stakeholders to understand how changes in LULC affect woodland-based ecosystem services (ES) in the Mabalane landscape, southern Mozambique. Three scenarios of policy interventions were tested: Large private; Small holder and Balanced. A BBNs was used to explore the influence of these scenarios from 2014 to 2035 on the resulting LULC. This research facilitated stakeholder engagement and improved the understanding of the interaction between LULC changes and woodland-based ES. The results highlighted the importance and spatial distribution of woodland-based ES to the local communities and that availability of suitable wood for ES will decrease under the first scenario.

How Suitable is Quantile Mapping For Postprocessing GCM Precipitation Forecasts?

GCMs are used by many national weather services to produce seasonal outlooks of atmospheric and oceanic conditions and fluxes. Postprocessing is often a necessary step before GCM forecasts can be applied in practice. Quantile mapping (QM) is rapidly becoming the method of choice by operational agencies to postprocess raw GCM outputs. The authors investigate whether QM is appropriate for this task. Ensemble forecast postprocessing methods should aim to 1) correct bias, 2) ensure forecasts are reliable in ensemble spread, and 3) guarantee forecasts are at least as skillful as climatology, a property called “coherence.” This study evaluates the effectiveness of QM in achieving these aims by applying it to precipitation forecasts from the POAMA model. It is shown that while QM is highly effective in correcting bias, it cannot ensure reliability in forecast ensemble spread or guarantee coherence. This is because QM ignores the correlation between raw ensemble forecasts and observations. When raw forecasts are not significantly positively correlated with observations, QM tends to produce negatively skillful forecasts. Even when there is significant positive correlation, QM cannot ensure reliability and coherence for postprocessed forecasts. Therefore, QM is not a fully satisfactory method for postprocessing forecasts where the issues of bias, reliability, and coherence pre-exist. Alternative postprocessing methods based on ensemble model output statistics (EMOS) are available that achieve not only unbiased but also reliable and coherent forecasts. This is shown with one such alternative, the Bayesian joint probability modeling approach.

Uncertainty analysis in multiscale modeling of concrete based on continuum micromechanics

Abstract Reliable models for structural materials are a vital issue. Hierarchical multiscale models for concrete, incorporating microstructural properties of the material, have been developed to predict the mechanical behavior. The present paper aims to study the major sources of uncertainty in the framework of multiscale modeling based on continuum micromechanics. The effect of uncertainy in input parameters on the Young's modulus of cement-based materials is studied systematically by means of a probabilistic multiscale modeling approach covering three length scales. Sensitivity and uncertainty analyses are adapted to assess the stochastic predictions of the multiscale model. The total order sensitivity indices are computed using the variance-based sensitivity analysis. The total uncertainty according to the adjustment factor approach, comprising the parameter uncertainty and the model uncertainty, is quantified, aiming at the comparison of two commonly utilized hydration models. Application of the probabilistic methodology reveals that the uncertainties of the model responses increase significantly during the upscaling process. Hence, considering the stochastic variability of the predicted mechanical properties is of utmost importance for the assessment and evaluation of the model responses.

Neural Semantic Personalized Ranking for item cold-start recommendation

Recommender systems help users deal with information overload and enjoy a personalized experience on the Web. One of the main challenges in these systems is the item cold-start problem which is very common in practice since modern online platforms have thousands of new items published every day. Furthermore, in many real-world scenarios, the item recommendation tasks are based on users' implicit preference feedback such as whether a user has interacted with an item. To address the above challenges, we propose a probabilistic modeling approach called Neural Semantic Personalized Ranking (NSPR) to unify the strengths of deep neural network and pairwise learning. Specifically, NSPR tightly couples a latent factor model with a deep neural network to learn a robust feature representation from both implicit feedback and item content, consequently allowing our model to generalize to unseen items. We demonstrate NSPR's versatility to integrate various pairwise probability functions and propose two variants based on the Logistic and Probit functions. We conduct a comprehensive set of experiments on two real-world public datasets and demonstrate that NSPR significantly outperforms the state-of-the-art baselines.

Evaluating the impacts of agricultural land management practices on water resources: A probabilistic hydrologic modeling approach

Evaluating the effectiveness of agricultural land management practices in minimizing environmental impacts using models is challenged by the presence of inherent uncertainties during the model development stage. One issue faced during the model development stage is the uncertainty involved in model parameterization. Using a single optimized set of parameters (one snapshot) to represent baseline conditions of the system limits the applicability and robustness of the model to properly represent future or alternative scenarios. The objective of this study was to develop a framework that facilitates model parameter selection while evaluating uncertainty to assess the impacts of land management practices at the watershed scale. The model framework was applied to the Lake Creek watershed located in southwestern Oklahoma, USA. A two-step probabilistic approach was implemented to parameterize the Agricultural Policy/Environmental eXtender (APEX) model using global uncertainty and sensitivity analysis to estimate the full spectrum of total monthly water yield (WYLD) and total monthly Nitrogen loads (N) in the watershed under different land management practices. Twenty-seven models were found to represent the baseline scenario in which uncertainty of up to 29% and 400% in WYLD and N, respectively, is plausible. Changing the land cover to pasture manifested the highest decrease in N to up to 30% for a full pasture coverage while changing to full winter wheat cover can increase the N up to 11%. The methodology developed in this study was able to quantify the full spectrum of system responses, the uncertainty associated with them, and the most important parameters that drive their variability. Results from this study can be used to develop strategic decisions on the risks and tradeoffs associated with different management alternatives that aim to increase productivity while also minimizing their environmental impacts.

A probabilistic model of uncertainties in the substructures and interfaces of a dynamical system: application to the torsional vibration of a drill-string

In this paper, a new strategy for modeling uncertainties in the substructures and interfaces of a dynamical system is presented. This strategy is based on (1) the reduction in the dynamical model of each substructure using the Craig–Bampton method and (2) the use of the nonparametric probabilistic approach for the global modeling of uncertainties in each substructure. As an improvement with respect to existing nonparametric methods, the methodology proposed here constructs separated models of uncertainties for the inner and interface degrees of freedom, which allows to control separately the levels of fluctuation induced by these two sources of uncertainties. This methodology is applied for the analysis of the random vibration of a drill-string. Three strategies are compared: (1) a full nonparametric probabilistic approach on all the system, (2) the existing nonparametric probabilistic approach together with the Craig–Bampton substructuring method, and (3) the new nonparametric probabilistic approach proposed here with the separation of the inner and interface degrees of freedom uncertainties. It turns out that, for the same level of uncertainty, the three approaches give similar results, but the new approach gives more flexibility for the control of the probabilistic model.

Mapping complementary features of cross-species structural connectivity to construct realistic "Virtual Brains".

Modern systems neuroscience increasingly leans on large-scale multi-lab neuroinformatics initiatives to provide necessary capacity for biologically realistic modeling of primate whole-brain activity. Here, we present a framework to assemble primate brain's biologically plausible anatomical backbone for such modeling initiatives. In this framework, structural connectivity is determined by adding complementary information from invasive macaque axonal tract tracing and non-invasive human diffusion tensor imaging. Both modalities are combined by means of available interspecies registration tools and a newly developed Bayesian probabilistic modeling approach to extract common connectivity evidence. We demonstrate how this novel framework is embedded in the whole-brain simulation platform called The Virtual Brain (TVB). Hum Brain Mapp 38:2080-2093, 2017. © 2017 Wiley Periodicals, Inc.

The Performance of Structure-Controller Coupled Systems Analysis Using Probabilistic Evaluation and Identification Model Approach

This study evaluates the performance of passively controlled steel frame building under dynamic loads using time series analysis. A novel application is utilized for the time and frequency domains evaluation to analyze the behavior of controlling systems. In addition, the autoregressive moving average (ARMA) neural networks are employed to identify the performance of the controller system. Three passive vibration control devices are utilized in this study, namely, tuned mass damper (TMD), tuned liquid damper (TLD), and tuned liquid column damper (TLCD). The results show that the TMD control system is a more reliable controller than TLD and TLCD systems in terms of vibration mitigation. The probabilistic evaluation and identification model showed that the probability analysis and ARMA neural network model are suitable to evaluate and predict the response of coupled building-controller systems.

Human-in-the-loop: application of the double exponential probability distribution function enables one to quantify the role of the human factor

The probabilistic predictive modelling approach in human-in-the-loop related aerospace problems enables one to predict, quantify, assure and even specify the probability of the favourable outcome of an aerospace mission or a situation, when the performance of the never perfect human, never 100%-reliable instrumentation (equipment), never absolutely predictable response of the object of control (aero- or space-craft), uncertain and often harsh environment, as well as the interaction (interfaces) of the above uncertainties, contribute jointly to the likelihood of such an outcome. As to the human factor, it includes two major aspects: human performance (error) and his/her state of health. While the reliability of the navigation instrumentation (equipment) could be evaluated using suitable and more or less well established modelling means, the role of the human factor, when quantification of the human role is critical, could be considered by using the double-exponential-probability-distribution-function (DEPDF).

Probabilistic Modeling of Single and Concurrent Truckloads on Bridges

The maintenance of bridges and the evolution of an appropriate bridge rating system require the consideration of loads from heavy trucks. These loads can arise from a single overweight truck or multiple trucks simultaneously present, or concurrent, on a bridge. This paper presents a probabilistic modeling approach to assess the frequency and likelihood of observing various bridge loads caused by single and concurrent trucks. The approach used weigh-in-motion (WIM) data collected at or near bridges of interest to identify single and concurrent trucks and their Corresponding loads. The modeling approach was applied to bridges near three WIM stations in Florida. Results showed that in any given month, there was a 100% probability of observing at least one single or concurrent truckload that exceeded twice the minimum weight of a single overweight truck (i.e., exceeded 711,715 N or 160,000 lb). In addition, the probability of observing extreme truckloads was significantly higher when all trucks were considered, as opposed to only overweight trucks. The modeling approach can easily be adapted to the goals of the study and to any region where WIM data are available at or near the bridge(s) of interest. Results generated from the modeling approach provide probabilistic loading input for bridge maintenance planning and truck permitting policy.