Bayesian Artificial Neural Network Research Articles

Bayesian Artificial Neural Networks for frontier efficiency analysis

Artificial neural networks have offered their share of econometric insights, given their power to model complex relationships. One area where they have not been readily deployed is the estimation of frontiers. The literature on frontier estimation has seen its share of research comparing and contrasting data envelopment analysis (DEA) and stochastic frontier analysis (SFA), the two workhorse estimators. These studies rely on both Monte Carlo experiments and actual data sets to examine a range of performance issues which can be used to elucidate insights on the benefits or weaknesses of one method over the other. As can be imagined, neither method is universally better than the other. The present paper proposes an alternative approach that is quite flexible in terms of functional form and distributional assumptions and it amalgamates the benefits of both DEA and SFA. Specifically, we bridge these two popular approaches via Bayesian artificial neural networks while accounting for possible endogeneity of inputs. We examine the performance of this new machine learning approach using Monte Carlo experiments which is found to be very good, comparable to, or often better than, the current standards in the literature. To illustrate the new techniques, we provide an application of this approach to a data set of large US banks.

Climate change impact assessment on groundwater level changes: A study of hybrid model techniques

AbstractOne of the most important sources of water supply is groundwater. However, the groundwater level (GWL) is significantly impacted by the global climate change. Therefore, under these more severe climate change conditions, the accurate and simple forecast of farmland GWL is a crucial component of agricultural water management. A hybrid model (HM) of Bayesian random forest (BRF), Bayesian support vector machine (BSVM), and Bayesian artificial neural network (BANN) is built in this study. The HM is made up of a Bayesian model averaging (BMA) and three machine learning models: random forest (RF), support vector machine (SVM), and artificial neural network. These three HMs are employed to help automate logical inference and decision‐making in business intelligence for groundwater management. For this purpose, data on 8 separate climatic factors that impact GWL changes in the study area were acquired. Nine distinct farming communities' GWL change data were utilised as the dependent variables for each model fit (community data). The effectiveness of the HM techniques was assessed using the evaluation metrics of mean absolute error (MAE), coefficient of determination (R2), mean absolute percent error (MAPE), and root mean square error (RMSE). The model fit in Suhum had the greatest performance with the highest accuracy (R2 varied from 0.9051 to 0.9679) and the lowest error scores (RMSE ranged from 0.0653 to 0.0727, and MAE ranged from 0.0121 to 0.0541), according to the models' evaluation results. The BRF delivered the greatest results when compared to the two independent HMs, the BSVM and BANN. Future GWL and climatic variable data may be trained using the trained HM techniques to determine the effects of climate change. Farmers, businesses, and civil society organisations might benefit from continuous monitoring of GWL data and education on climate change to help control and prevent excessive deteriorations of global climate change on GWL.

A new method for modeling radiative heat transfer based on Bayesian artificial neural networks and Monte Carlo method in participating media

A new method based on predictive capacity of Feedforward Artificial Neural Networks (FANN) is proposed to estimate the divergence of the radiative flux in an axisymmetric domain. Training and validation databases have been built thanks to results given by the SNB-CK model and computed accordingly with a null collision Monte Carlo algorithm. The major aim of this work is to combine advantages of spectral models in terms of accuracy and the computational efficiency of neural networks in order to make possible the accurate modeling of radiative heat transfer. As a result, ANNs are able to model the radiative flux divergence on the basis of training data and some keys to avoid the pitfalls related to ANNs are provided.

A machine-learning algorithm for distinguishing malignant from benign indeterminate thyroid nodules using ultrasound radiomic features.

Background: Ultrasound (US)-guided fine needle aspiration (FNA) cytology is the gold standard for the evaluation of thyroid nodules. However, up to 30% of FNA results are indeterminate, requiring further testing. In this study, we present a machine-learning analysis of indeterminate thyroid nodules on ultrasound with the aim to improve cancer diagnosis. Methods: Ultrasound images were collected from two institutions and labeled according to their FNA (F) and surgical pathology (S) diagnoses [malignant (M), benign (B), and indeterminate (I)]. Subgroup breakdown (FS) included: 90 BB, 83 IB, 70 MM, and 59 IM thyroid nodules. Margins of thyroid nodules were manually annotated, and computerized radiomic texture analysis was conducted within tumor contours. Initial investigation was conducted using five-fold cross-validation paradigm with a two-class Bayesian artificial neural networks classifier, including stepwise feature selection. Testing was conducted on an independent set and compared with a commercial molecular testing platform. Performance was evaluated using receiver operating characteristic analysis in the task of distinguishing between malignant and benign nodules. Results: About 1052 ultrasound images from 302 thyroid nodules were used for radiomic feature extraction and analysis. On the training/validation set comprising 263 nodules, five-fold cross-validation yielded area under curves (AUCs) of 0.75 [Standard Error (SE) = 0.04; ] and 0.67 (SE = 0.05; ) for the classification tasks of MM versus BB, and IM versus IB, respectively. On an independent test set of 19 IM/IB cases, the algorithm for distinguishing indeterminate nodules yielded an AUC value of 0.88 (SE = 0.09; ), which was higher than the AUC of a commercially available molecular testing platform (AUC = 0.81, SE = 0.11; ). Conclusion: Machine learning of computer-extracted texture features on gray-scale ultrasound images showed promising results classifying indeterminate thyroid nodules according to their surgical pathology.

Wind Turbine Power Curve Modelling using Gaussian Mixture Copula, ANN Regressive and BANN

In terms of monitoring wind turbine performance, the power curve plays a key role and it is a valuable indicator of the wind turbine’s condition as it allows the identification of anomalies related to production. On the other hand, by calculating a typical reference curve of the machine, it is possible to calculate losses and estimate its performance. A suitable model that characterises the power curve can benefit the estimated production of a wind turbine or wind farm. The most widely used models to characterise power curves are based on 2-variable models (wind speed and power). The power curve is affected by different exogenous variables (such as temperature, wind direction or turbulence intensity) that must be taken into account in the characterisation of the power curve. Consequently, the working conditions of the wind turbine are different depending on the site, and multivariable models that can take these variables into account are required. Currently new models and alternatives are emerging. This study will evaluate the performance of multivariate models based on Gaussian mixture copula model (GMCM), Regressive artificial neural Networks (ANN) and Bayesian artificial neural networks (BANN) in terms of power curve characterisation and power estimation. The results will be compared to the most common methods such as the manufacturer power curve and historical power curve, which is generated using the the bins method. Taking advantage of the benefits of neural networks (NN), particular types of NN have been considered to provide the value of uncertainty in the estimation of power.

Modeling Semiarid River–Aquifer Systems with Bayesian Networks and Artificial Neural Networks

In semiarid areas, precipitations usually appear in the form of big and brief floods, which affect the aquifer through water infiltration, causing groundwater temperature changes. These changes may have an impact on the physical, chemical and biological processes of the aquifer and, thus, modeling the groundwater temperature variations associated with stormy precipitation episodes is essential, especially since this kind of precipitation is becoming increasingly frequent in semiarid regions. In this paper, we compare the predictive performance of two popular tools in statistics and machine learning, namely Bayesian networks (BNs) and artificial neural networks (ANNs), in modeling groundwater temperature variation associated with precipitation events. More specifically, we trained a total of 2145 ANNs with different node configurations, from one to five layers. On the other hand, we trained three different BNs using different structure learning algorithms. We conclude that, while both tools are equivalent in terms of accuracy for predicting groundwater temperature drops, the computational cost associated with the estimation of Bayesian networks is significantly lower, and the resulting BN models are more versatile and allow a more detailed analysis.

Metaheuristic Approaches Integrated with ANN in Forecasting Daily Emergency Department Visits

The overall service quality level of Emergency Departments (EDs) can be improved by accurate forecasting of patient visits. Accordingly, this study aims to evaluate the use of three metaheuristic approaches integrated with Artificial Neural Network (ANN) in forecasting daily ED visits. To do this, five performance measures are used for evaluating the accuracy of the proposed approaches, including Bayesian ANN, Genetic Algorithm-based ANN (GA-ANN), and Particle Swarm Optimization algorithm-based ANN (PSO-ANN). The outputs of this study show that the PSO-ANN model provides the most dominant performance in both the training and testing process. The lowest error is obtained with a mean absolute percentage error (MAPE) of 6.3%, Mean Absolute Error (MAE) of 42.797, Mean Squared Error (MSE) of 2499.340, Root Mean Square Error (RMSE) of 49.933, and R-squared (R2) of 0.824 on the training dataset. The lowest error with an MAPE of 6.0%, MAE of 40.888, MSE of 2839.998, RMSE of 53.292, and R2 of 0.791 is also obtained on the testing process.

Elucidating the effect of process parameters on the production of hydrogen-rich syngas by biomass and coal Co-gasification techniques: A multi-criteria modeling approach

The thermochemical processes such as gasification and co-gasification of biomass and coal are promising route for producing hydrogen-rich syngas. However, the process is characterized with complex reactions that pose a tremendous challenge in terms of controlling the process variables. This challenge can be overcome using appropriate machine learning algorithm to model the nonlinear complex relationship between the predictors and the targeted response. Hence, this study aimed to employ various machine learning algorithms such as regression models, support vector machine regression (SVM), gaussian processing regression (GPR), and artificial neural networks (ANN) for modeling hydrogen-rich syngas production by gasification and co-gasification of biomass and coal. A total of 12 machine learning algorithms which comprises the regression models, SVM, GPR, and ANN were configured, trained using 124 datasets. The performances of the algorithms were evaluated using the coefficient of determination (R2), root mean square error (RMSE), mean square error (MSE), and mean absolute error (MAE). In all cases, the ANN algorithms offer superior performances and displayed robust predictions of the hydrogen-rich syngas from the co-gasification processes. The R2 of both the Levenberg-Marquardt- and Bayesian Regularization-trained ANN obtained from the prediction of the hydrogen-rich syngas was found to be within 0.857–0.998 with low prediction errors. The sensitivity analysis to determine the effect of the process parameters on the model output revealed that all the parameters showed a varying level of influence. In most of the processes, the gasification temperature was found to have the most significant influence on the model output.

Artificial neural networks and phenomenological degradation models for fatigue damage tracking and life prediction in laser induced graphene interlayered fiberglass composites

The mechanical properties of fiber reinforced polymer matrix composites are known to gradually deteriorate as fatigue damage accumulates under cyclic loading conditions. While the steady degradation in elastic stiffness throughout fatigue life is a well-established and studied concept, it remains difficult to continuously monitor such structural changes during the service life of many dynamic engineering systems where composite materials are subjected to random and unexpected loading conditions. Recently, laser induced graphene (LIG) has been demonstrated to be a reliable, in-situ strain sensing and damage detection component in fiberglass composites under both quasi-static and dynamic loading conditions. This work investigates the potential of exploiting the piezoresistive properties of LIG interlayered fiberglass composites in order to formulate cumulative damage parameters and predict both damage progression and fatigue life using artificial neural networks (ANNs) and conventional phenomenological models. The LIG interlayered fiberglass composites are subjected to tension–tension fatigue loading, while changes in their elastic stiffness and electrical resistance are monitored through passive measurements. Damage parameters that are defined according to changes in electrical resistance are found to be capable of accurately describing damage progression in LIG interlayered fiberglass composites throughout fatigue life, as they display similar trends to those based on changes in elastic stiffness. These damage parameters are then exploited for predicting the fatigue life and future damage state of fiberglass composites using both trained ANNs and phenomenological degradation and accumulation models in both specimen-to-specimen and cycle-to-cycle schemes. When used in a specimen-to-specimen scheme, the predictions of a two-layer Bayesian regularized ANN with 40 neurons in each layer are found to be at least 60% more accurate than those of phenomenological degradation models, displaying R 2 values greater than 0.98 and root mean square error (RMSE) values smaller than 10−3. A two-layer Bayesian regularized ANN with 25 neurons in each layer is also found to yield accurate predictions when used in a cycle-to-cycle scheme, displaying R 2 values greater than 0.99 and RMSE values smaller than 2 × 10−4 once more than 30% of the initial measurements are used as inputs. The final results confirm that piezoresistive LIG interlayers are a promising tool for achieving accurate and continuous fatigue life predictions in multifunctional composite structures, specifically when coupled with machine learning algorithms such as ANNs.

Comparative Analysis of Artificial Intelligence Models for Accurate Estimation of Groundwater Nitrate Concentration.

Prediction of the groundwater nitrate concentration is of utmost importance for pollution control and water resource management. This research aims to model the spatial groundwater nitrate concentration in the Marvdasht watershed, Iran, based on several artificial intelligence methods of support vector machine (SVM), Cubist, random forest (RF), and Bayesian artificial neural network (Baysia-ANN) machine learning models. For this purpose, 11 independent variables affecting groundwater nitrate changes include elevation, slope, plan curvature, profile curvature, rainfall, piezometric depth, distance from the river, distance from residential, Sodium (Na), Potassium (K), and topographic wetness index (TWI) in the study area were prepared. Nitrate levels were also measured in 67 wells and used as a dependent variable for modeling. Data were divided into two categories of training (70%) and testing (30%) for modeling. The evaluation criteria coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE), and Nash–Sutcliffe efficiency (NSE) were used to evaluate the performance of the models used. The results of modeling the susceptibility of groundwater nitrate concentration showed that the RF (R2 = 0.89, RMSE = 4.24, NSE = 0.87) model is better than the other Cubist (R2 = 0.87, RMSE = 5.18, NSE = 0.81), SVM (R2 = 0.74, RMSE = 6.07, NSE = 0.74), Bayesian-ANN (R2 = 0.79, RMSE = 5.91, NSE = 0.75) models. The results of groundwater nitrate concentration zoning in the study area showed that the northern parts of the case study have the highest amount of nitrate, which is higher in these agricultural areas than in other areas. The most important cause of nitrate pollution in these areas is agriculture activities and the use of groundwater to irrigate these crops and the wells close to agricultural areas, which has led to the indiscriminate use of chemical fertilizers by irrigation or rainwater of these fertilizers is washed and penetrates groundwater and pollutes the aquifer.

Modeling the effect of non‐linear process parameters on the prediction of hydrogen production by steam reforming of bio‐oil and glycerol using artificial neural network

Biomass-derived substrates such as bio-oil and glycerol are gaining wide acceptability as feedstocks to produce hydrogen using a steam reforming process. The wide acceptability can be attributed to a huge amount of glycerol and bio-oil obtained as by-products of biodiesel production and pyrolysis processes. Several parameters have been reported to affect the production of hydrogen by biomass steam reforming. This study investigates the effect of non-linear process parameters on the prediction of hydrogen production by biomass (bio-oil and glycerol) steam reforming using artificial neural network (ANN) modeling technique. Twenty different multilayer ANN model architectures were tested using datasets obtained from the bio-oil and glycerol steam reforming. Two algorithms namely Levenberg-Marquardt and Bayesian regularization were employed for the training of the ANNs. An optimized network configuration consisting of 3 input layer 14 hidden neurons, 1 output layer, and 3 input layer, 5 hidden neurons, and 1 output layer were obtained for the Levenberg-Marquardt and Bayesian regularization trained network, respectively for hydrogen production by bio-oil steam reforming. While an optimized network configuration consisting of 5 input nodes, 9 hidden neurons, 1 output node, and 5 input nodes, 8 hidden neurons, and 1 output node were obtained for Levenberg-Marquardt and Bayesian regularization trained network, respectively for hydrogen production by glycerol steam reforming. Based on the optimized network, the predicted hydrogen production from the bio-oil and glycerol steam agreed with the actual values with the coefficient of determination (R2) > 0.9. A low mean square error of 3.024 × 10−24 and 6.22 × 10−15 for the optimized for Levenberg-Marquardt and Bayesian regularization-trained ANN, respectively. The neural network analyses of the two processes showed that reaction temperature and glycerol-to-water molar ratio were the most relevant factors that influenced the production of hydrogen by bio-oil and glycerol steam reforming, respectively. This study has demonstrated the robustness of the ANN as a technique for investigating the effect of non-linear process parameters on hydrogen production by bio-oil and glycerol steam reforming.

Influence of Segregation on Microstructure and Hot Workability of Grade 250 Maraging Steel

The optimal process parameters of grade 250 maraging steel for hot forging were investigated with a process map and microstructure analysis. To generate the process map, hot compression tests were performed at 800–1200 °C and strain rates of 0.01–5 s−1. The flow stress–strain curves were calibrated by Bayesian artificial neural network (ANN) modeling to compensate the heat generated by dynamic deformation. The Ni and Mo segregation during the solidification of the ingot caused alternating layers of Ni- and Mo-rich and -lean bands, which affected the recrystallization behavior during hot compression. According to the calculated process map, 1100–1200 °C × 0.01–0.7 s−1 and 1000–1200 °C × 0.01–0.2 s−1 are favorable process conditions that ensure wide process windows in terms of strain rate and temperature, respectively. The common features of the microstructure deformed under both conditions were relatively coarse martensite blocks and low values in the electron backscattered diffraction (EBSD) kernel average misorientation (KAM) results (i.e., low residual stresses), which were attributed to a low fraction of fine martensite block region.

Моделирование изменения предела длительной прочности сплавов на основе никеля с использованием байесовских искусственных нейронных сетей

Моделирование изменения предела длительной прочности сплавов на основе никеля с использованием байесовских искусственных нейронных сетей

Physical-empirical models for prediction of seasonal rainfall extremes of Peninsular Malaysia

Abstract Reliable prediction of rainfall extremes is vital for disaster management, particularly in the context of increasing rainfall extremes due to global climate change. Physical-empirical models have been developed in this study using three widely used Machine Learning (ML) methods namely, Support Vector Machines (SVM), Random Forests (RF), Bayesian Artificial Neural Networks (BANN) for the prediction of rainfall and rainfall related extremes during Northeast Monsoon (NEM) in Peninsular Malaysia from synoptic predictors. The gridded daily rainfall data of Asian Precipitation—Highly Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) was used to estimate four rainfall indices namely, rainfall amount, average rainfall intensity, days having >95-th percentile rainfall, and total number of dry days in Peninsular Malaysia during NEM for the period 1951–2015. The National Centers for Environmental Prediction (NCEP) reanalysis sea level pressure (SLP) data was used for the prediction of rainfall indices with different lead periods. The recursive feature elimination (RFE) method was used to select the SLP at different NCEP grid points which were found significantly correlated with NEM rainfall indices. The results showed superior performance of BANN among the ML models with normalised root mean square error of 0.04–0.14, Nash-Sutcliff Efficiency of 0.98–1.0, and modified agreement index of 0.97–0.99 and Kling-Gupta efficient index 0.65–0.96 for one-month lead period prediction. The 95% confidence interval (CI) band for BANN was found narrower than the other ML models. Almost all the forecasted values by BANN were also found with 95% CI, and therefore, the p-factor and the r-factor for BANN in predicting rainfall indices were found in the range of 0.95–1.0 and 0.25–0.49 respectively. Application of BANN in prediction of rainfall indices with higher lead time was also found excellent. The synoptic pattern revealed that SLP over the north of South China Sea is the major driver of NEM rainfall and rainfall extremes in Peninsular Malaysia.

Bayesian modelling of nonlinear Poisson regression with artificial neural networks

ABSTRACT Modelling and prediction of count and rate responses have substantial usage in many fields, including health, finance, social, etc. Conventionally, linear Poisson regression models have been widely used to model these responses. However, the linearity assumption of the systematic component of linear Poisson regression models restricts their capability of handling complex data patterns. In this regard, it is important to develop nonlinear Poisson regression models to capture the inherent variability within the count data. In this study, we introduce a probabilistically driven nonlinear Poisson regression model with Bayesian artificial neural networks (ANN) to model count and rate data. This new nonlinear Poisson regression model developed with Bayesian ANN provides higher prediction accuracies over traditional Poisson or negative binomial regression models as revealed in our simulation and real data studies.

Developing comparative mathematic models, BN and ANN for forecasting of groundwater levels

An accurate prediction of groundwater level is a key element in appropriate planning and managing benefits driven from aquifers in watershed. In this study, the conceptual model (MODFLOW), the Bayesian network (BN) and the artificial neural network (ANN) models are employed to forecast monthly groundwater level in Birjand Aquifer, South Khorasan, Iran. In this regard, total monthly evaporation, average temperature, aquifer recharge and discharge, and water table of previous months from 13 observation piezometers and in a 12-year period are used as input variables in order to forecast the groundwater level in the forthcoming months. The results of different models revealed that the Bayesian network models are superior to the ANN and mathematical models. The Bayesian network, ANN and mathematical models developed for the 13 piezometers have an average coefficient of determination of 0.9, 0.76 and 0.72, respectively. The results of this study demonstrate that the BNs are efficient tools for forecasting groundwater level.

Dynamic modeling of room temperature and thermodynamic efficiency for direct expansion air conditioning systems using Bayesian neural network

In this paper, dynamic performance identification for a direct expansion (DX) air conditioning (AC) system is proposed using Bayesian artificial neural network (ANN). The input and output datasets are generated by a dedicated AC simulator by varying the compressor speed in various signal amplitudes and including dynamic cooling load and ambient temperature. The exergy destruction, which represents the work potential losses in the system and room temperature indicating the thermal comfort are selected as the output variables. The key parameters of an ANN model, including the number of neurons and tapped delay lines, are optimized to improve the prediction accuracy. The results show that the dynamic response of the exergy destruction and room temperature can be predicted accurately by the optimized ANN model using three neurons, a Bayesian regularization algorithm, five delayed inputs for the compressor speed and room temperature, and six delayed inputs for the cooling load and ambient temperature. The validation of the multi-step-ahead prediction showed satisfying results with respect to the root mean squared errors (RMSEs) and coefficient of variations (CVs) of the room temperature (RMSE: 0.18 °C and CV: 0.85%) and exergy destruction (RMSE: 1.79 W and CV: 0.4%). Accordingly, the identification of the AC system behavior presented in this paper could be further implemented to control the DX AC system operation to achieve a desired thermal comfort with low exergy destruction.

MADE: a spectroscopic mass, age, and distance estimator for red giant stars with Bayesian machine learning

We present a new approach (MADE) that generates mass, age, and distance estimates of red giant stars from a combination of astrometric, photometric, and spectroscopic data. The core of the approach is a Bayesian artificial neural network (ANN) that learns from and completely replaces stellar isochrones. The ANN is trained using a sample of red giant stars with mass estimates from asteroseismology. A Bayesian isochrone pipeline uses the astrometric, photometric, spectroscopic, and asteroseismology data to determine posterior distributions for the training outputs: mass, age, and distance. Given new inputs, posterior predictive distributions for the outputs are computed, taking into account both input uncertainties, and uncertainties in the ANN parameters. We apply MADE to $\sim10\,000$ red giants in the overlap between the 14$^{\mathrm{th}}$ data release from the APO Galactic Evolution Experiment (APOGEE, Abolfathi et al. 2018) and the Tycho-Gaia astrometric solution (TGAS, Michalik et al. 2015). The ANN is able to reduce the uncertainty on mass, age, and distance estimates for training-set stars with high output uncertainties allocated through the Bayesian isochrone pipeline. The fractional uncertainties on mass are $<10\%$ and on age are between $10$ to $25\%$. Moreover, the time taken for our ANN to predict masses, ages, and distances for the entire catalogue of APOGEE-TGAS stars is of a similar order of the time taken by the Bayesian isochrone pipeline to run on a handful of stars. Our resulting catalogue clearly demonstrates the expected thick and thin disc components in the [M/H]-[$\alpha$/M] plane, when examined by age.

Forewarning of hypotensive events using a Bayesian artificial neural network in neurocritical care.

Traumatically brain injured (TBI) patients are at risk from secondary insults. Arterial hypotension, critically low blood pressure, is one of the most dangerous secondary insults and is related to poor outcome in patients. The overall aim of this study was to get proof of the concept that advanced statistical techniques (machine learning) are methods that are able to provide early warning of impending hypotensive events before they occur during neuro-critical care. A Bayesian artificial neural network (BANN) model predicting episodes of hypotension was developed using data from 104 patients selected from the BrainIT multi-center database. Arterial hypotension events were recorded and defined using the Edinburgh University Secondary Insult Grades (EUSIG) physiological adverse event scoring system. The BANN was trained on a random selection of 50% of the available patients (n = 52) and validated on the remaining cohort. A multi-center prospective pilot study (Phase 1, n = 30) was then conducted with the system running live in the clinical environment, followed by a second validation pilot study (Phase 2, n = 49). From these prospectively collected data, a final evaluation study was done on 69 of these patients with 10 patients excluded from the Phase 2 study because of insufficient or invalid data. Each data collection phase was a prospective non-interventional observational study conducted in a live clinical setting to test the data collection systems and the model performance. No prediction information was available to the clinical teams during a patient's stay in the ICU. The final cohort (n = 69), using a decision threshold of 0.4, and including false positive checks, gave a sensitivity of 39.3% (95% CI 32.9-46.1) and a specificity of 91.5% (95% CI 89.0-93.7). Using a decision threshold of 0.3, and false positive correction, gave a sensitivity of 46.6% (95% CI 40.1-53.2) and specificity of 85.6% (95% CI 82.3-88.8). With a decision threshold of 0.3, > 15min warning of patient instability can be achieved. We have shown, using advanced machine learning techniques running in a live neuro-critical care environment, that it would be possible to give neurointensive teams early warning of potential hypotensive events before they emerge, allowing closer monitoring and earlier clinical assessment in an attempt to prevent the onset of hypotension. The multi-centre clinical infrastructure developed to support the clinical studies provides a solid base for further collaborative research on data quality, false positive correction and the display of early warning data in a clinical setting.

Comparison of Deformation Models of Flexible Manipulator Joints for Use in DEMO

The joint deformation contributes significantly to final end-effector displacement of a manipulator, especially when the manipulator is in the form a serial kinematics with long links. When the manipulators employed in the DEMO are under the heavy payload, the deformation of manipulator is inevitable and the magnitude is significant. In order to maneuver the large object through several via points with high positioning accuracy in the remote handling process of DEMO, the real-time computation of manipulator deformation has to be conducted, which is crucial to the DEMO adaptive position control system for the displacement compensation. Three computation-effective deformation modeling methods are proposed in this paper, which are parametric modeling method, nonparametric deterministic artificial neural network (ANN) modeling method, and nonparametric Bayesian ANN modeling method, respectively. A specific joint in a boom equipped in a telescopic articulated remote mast is taken as the study object in this paper. A nodal deformation in the joint is investigated by three modeling methods, respectively. The parametric deformation model is derived by using the structural mechanics, whose parameters are identified by using the Markov chain Monte Carlo (MCMC) method; the deformation model of deterministic ANN is trained by using the Levenberg–Marquardt method; and the deformation model of the Bayesian ANN is trained by using the MCMC method. The results show that the parametric model from the structural mechanics is linear and is incompetent in the deformation modeling when the nonlinearity presents; both the deterministic and Bayesian ANNs are capable of model the nodal deformation of joint. The performance of both the deterministic network and Bayesian network cannot rival for one another in the application scenario of this paper. The training of the Bayesian network can provide the criterions for estimation of possible ranges of the modeling outputs from its probabilistic distribution curves, and the judgment of proper size of network.