Linear Network Model Research Articles

Simple and Rapid Discrimination of Methicillin-Resistant Staphylococcus aureus Based on Gram Staining and Machine Vision.

Methicillin-resistant Staphylococcus aureus (MRSA) is a clinical threat with high morbidity and mortality. Here, we describe a new simple, rapid identification method for MRSA using oxacillin sodium salt, a cell wall synthesis inhibitor, combined with Gram staining and machine vision (MV) analysis. Gram staining classifies bacteria as positive (purple) or negative (pink) according to the cell wall structure and chemical composition. In the presence of oxacillin, the integrity of the cell wall for methicillin-susceptible S. aureus (MSSA) was destroyed immediately and appeared Gram negative. In contrast, MRSA was relatively stable and appeared Gram positive. This color change can be detected by MV. The feasibility of this method was demonstrated in 150 images of the staining results for 50 clinical S. aureus strains. Based on effective feature extraction and machine learning, the accuracies of the linear linear discriminant analysis (LDA) model and nonlinear artificial neural network (ANN) model for MRSA identification were 96.7% and 97.3%, respectively. Combined with MV analysis, this simple strategy improved the detection efficiency and significantly shortened the time needed to detect antibiotic resistance. The whole process can be completed within 1 h. Unlike the traditional antibiotic susceptibility test, overnight incubation is avoided. This new strategy could be used for other bacteria and represents a new rapid method for detection of clinical antibiotic resistance. IMPORTANCE Oxacillin sodium salt destroys the integrity of the cell wall of MSSA immediately, appearing Gram negative, whereas MRSA is relatively stable and still appears Gram positive. This color change can be detected by microscopic examination and MV analysis. This new strategy has significantly reduced the time to detect resistance. The results show that using oxacillin sodium salt combined with Gram staining and MV analysis is a new, simple and rapid method for identification of MRSA.

Trade-offs between aggregated and turbine-level representations of hydropower in optimization models

To model a future power system with high shares of variable renewables, it is essential to capture the flexibility of dispatchable technologies such as hydropower. However, the representation of hydropower is often oversimplified in energy system investment models, such that the flexibility of hydropower is significantly exaggerated. This suggests the need for improved representations of hydropower that capture physical river dynamics but are computationally efficient to maintain the tractability of large models. Here, we develop a series of hydropower optimization models for a single river with various levels of techno-physical detail to evaluate options for hydropower representations in energy system investment models. All models operate hourly over a full year with perfect foresight. We explore trade-offs between accuracy and computational time involved in including features such as the river network, head-dependent power production, and discharge-dependent turbine efficiencies. We find that the level of detail significantly affects the optimal production and confirm that a simplistic hydropower representation similar to those often used in investment models significantly overestimates the flexibility of hydropower. The most detailed nonconvex model includes a full river network, head-dependency, and turbine efficiencies and is solved in just one hour on a modern desktop computer. Furthermore, we linearize this detailed model, thereby reducing computation time to one minute while featuring production dynamics substantially more similar to the full nonconvex model than a naive linear network model. These contributions pave the way for improving hydropower representations in investment models to avoid overestimating the flexibility that hydropower may provide.

A Generalization Gap Estimation for Overparameterized Models via the Langevin Functional Variance

This article discusses the estimation of the generalization gap, the difference between generalization performance and training performance, for overparameterized models including neural networks. We first show that a functional variance, a key concept in defining a widely-applicable information criterion, characterizes the generalization gap even in overparameterized settings where a conventional theory cannot be applied. As the computational cost of the functional variance is expensive for the overparameterized models, we propose an efficient approximation of the function variance, the Langevin approximation of the functional variance (Langevin FV). This method leverages only the first-order gradient of the squared loss function, without referencing the second-order gradient; this ensures that the computation is efficient and the implementation is consistent with gradient-based optimization algorithms. We demonstrate the Langevin FV numerically by estimating the generalization gaps of overparameterized linear regression and nonlinear neural network models, containing more than a thousand of parameters therein. Supplementary materials for this article are available online.

Accelerating distributed machine learning with model compression and graph partition

The rapid growth of data and parameter sizes of machine learning models makes it necessary to improve the efficiency of distributed training. It is observed that the communication cost usually is the bottleneck of distributed training systems. In this paper, we focus on the parameter server framework which is a widely deployed distributed learning framework. The frequent parameter pull, push, and synchronization among multiple machines leads to a huge communication volume. We aim to reduce the communication cost for the parameter server framework. Compressing the training model and optimizing the data and parameter allocation are two existing approaches to reducing communication costs. We jointly consider these two approaches and propose to optimize the data and parameter allocation after compression. Different from previous allocation schemes, the data sparsity property may no longer hold after compression. It brings additional opportunities and challenges for the allocation problem. We also consider the allocation problem for both linear and deep neural network (DNN) models. Fixed and dynamic partition algorithms are proposed accordingly. Experiments on real-world datasets show that our joint compression and partition scheme can efficiently reduce communication overhead for linear and DNN models.

Physics Informed Piecewise Linear Neural Networks for Process Optimization

Constructing first-principles models is usually a challenging and time-consuming task due to the complexity of real-life processes. On the other hand, data-driven modeling, particularly a neural network model, often suffers from overfitting and lack of useful and high-quality data. At the same time, embedding trained machine learning models directly into the optimization problems has become an effective and state-of-the-art approach for surrogate optimization, whose performance can be improved by physics-informed machine learning. This study proposes using piecewise linear neural network models with physics-informed knowledge for optimization problems with neural network models embedded. In addition to using widely accepted and naturally piecewise linear rectified linear unit (ReLU) activation functions, this study also suggests piecewise linear approximations for the hyperbolic tangent activation function to widen the domain. Optimization of three case studies, a blending process, an industrial distillation column, and a crude oil column are investigated. Physics-informed trained neural network-based optimal results are closer to global optimality for all cases. Finally, associated CPU times for the optimization problems are much shorter than the standard optimization results.

Social Networks with Unobserved Links

We point-identify and estimate linear social network models without observing any network links. The required data consist of many small networks of individuals, such as classrooms or villages, with individuals who are each observed only once. We apply our estimator to data from Tennessee’s Project STAR (Student-Teacher Achievement Ratio). Without observing the latent network in each classroom, we identify and estimate peer and contextual effects on students’ performance in mathematics. We find that peer effects tend to be larger in bigger classes and that increasing peer effects would significantly improve students’ average test scores in some classes.

Optimization of solid oxide fuel cell power generation voltage prediction based on improved neural network

Abstract This paper proposes a method for predicting the generation voltage of a solid oxide fuel cell based on the data results of a stand-alone solid oxide fuel single cell simulation model under ideal conditions, with the aim of improving the generation efficiency and extending the service life of the solid oxide fuel cell. In this paper, a modified back propagation (BP) neural network algorithm is used to improve the prediction accuracy of the solid oxide fuel cell generation voltage by using the whale algorithm to optimize the BP neural network model to improve its convergence and achieve the effect of improving the prediction accuracy. First, the characteristics of the independent solid oxide fuel cell are introduced and simulated. Second, the long short-term memory network model, linear regression network model and BP neural network are simulated and compared, and the results show that the BP neural network prediction model is more accurate and can be optimized and improved. Finally, the BP neural network is optimized and simulated using the whale algorithm, and the simulation results show that the method has better convergence and higher prediction accuracy than the traditional BP neural network prediction model.

PREDICTION OF THE DYSBIOSIS DEVELOPMENT RISK ACCORDING TO HUMORAL IMMUNITY INDICATORS

Diseases associated with bacterial vaginosis lead to chronic inflammatory processes of the internal genitals, the development of adhesions of the pelvic organs, infertility, spontaneous abortion at different times, as well as the development of malignant neoplasms. Vaginal microflora is an indicator of a woman’s health, which can form changing in hormonal and immunological status during various pathological conditions. The aim of the study was to create a system for prediction of the dysbiosis development according to the levels of nonspecific humoral factor of immune defense. Materials and methods. The study was performed in 298 women aged 16 to 64 years, 53 of whom were diagnosed with normocenosis, and 245 have dysbiosis. Women were divided into 3 groups according to age. Regression analysis was used. Research results and their discussion. Our previous research have shown a correlation between increased levels of anti-inflammatory cytokines in the blood and vaginal secretions with the stage of dysbiosis. A logistic regression model was constructed during the study, which showed that the risk of developing dysbiosis in terms of normobiota increases with increasing levels of interleukin 2 in the blood, tumor necrosis factor α. Significant features of the three-factor model for predicting the risk of developing dysbiosis (IL2, IL4 and TNFα) were selected by the method of genetic algorithm. The levels of these indicators in the blood were related to the severity of dysbiosis according to the results of discriminant analysis. Conclusions. Thus, a linear neural network model was developed for determination of dysbiosis severity according to the levels of nonspecific humoral factors of immune defense such as the C4 component of the complement system and γ-interferon in vaginal secretions, as well as the amount of circulating immune complexes and tumor necrosis factor α in the blood. Kappa Cohen’s agreement for this model on the training set was 0.87 (95% CI 0.82-0.91), and on the confirmatory set was 0.89 (95% CI 0.77-1.00). These indicators show the adequacy of the constructed model. The interface of the expert system for the dysbiosis severity prediction has been created. Key words: normobiota, bacterial vaginosis, linear neural network model for determination of the dysbiosis severity, humoral factor of immune defense.

Actuator Selection for Dynamical Networks with Multiplicative Noise

We propose a greedy algorithm for actuator selection considering multiplicative noise in the dynamics and actuator architecture of a discrete-time, linear system network model. We show that the resultant architecture achieves mean-square stability with lower control costs and for smaller actuator sets than the deterministic model, even in the case of modeling uncertainties. Networks with multiplicative noise may fail to be mean-square stabilizable by any small actuator set, leading to a failure of a cost-based greedy algorithm. To account for this, we propose a multi-metric greedy algorithm that allows actuator sets to be evaluated effectively even when none of them stabilize the system. We illustrate our results on networks with multiplicative noise in the open-loop dynamics and the actuator inputs, and we analyze control costs for random graphs of different network sizes and generation parameters.

Association of Multi-Dimensional Factors with Accelerating Age and Constructing a Healthy Lifestyle Index: Guangzhou Biobank Cohort Study

Introduction: Ageing process is influenced by multi-dimensional factors collectively. Previous studies examined association of one separate factor with mortality without considering different manifestations of ageing process. We investigated associations of multi-dimensional factors with accelerating age (AA), a proxy to quantify ageing, in older Chinese. Methods: 9,831 participants from Guangzhou Biobank Cohort Study were included. Four exposure domains of 15 variables including demographic and socio-economic factors, lifestyle factors, stress across the life course, and common diseases were assessed. AA was calculated based on chronological age and eight biomarkers. Traditional multivariable linear and Bayesian Network (BN) models were used. Results: In both traditional and BN models, male sex, smoking, alcohol use, physical inactivity, greater waist circumference, and body mass index (BMI) were associated with higher AA, with the adjusted β (95% confidence intervals) being 2.75 (2.40–3.09), 1.31 (0.87–1.76), 1.35 (0.55–2.15), 0.64 (0.40–0.88), 0.09 (0.06–0.11), and 0.13 (0.07–0.19) years, respectively. A Healthy Lifestyle Index (HLI) was constructed including the above lifestyle factors (non-smoking, non-alcohol use, physically active, non-central, and non-general obesity) with a point assigned for each. A higher index indicates healthier lifestyle. Compared with participants with an HLI of 5, those with an HLI of 0–2 had 2.90 (2.48–3.32) years older AA. Conclusions: Male sex, smoking, alcohol use, physical inactivity, greater waist circumference, and BMI were associated with higher AA by 0.09–2.75 years, suggesting that adopting a healthy lifestyle may alleviate process of phenotypic ageing.

A Combination of Near-Infrared Hyperspectral Imaging with Two-Dimensional Correlation Analysis for Monitoring the Content of Alanine in Beef.

Alanine (Ala), as the most important free amino acid, plays a significant role in food taste characteristics and human health regulation. The feasibility of using near-infrared hyperspectral imaging (NIR-HSI) combined with two-dimensional correlation spectroscopy (2D-COS) analysis to predict beef Ala content quickly and nondestructively is first proposed in this study. With Ala content as the external disturbance condition, the sequence of chemical bond changes caused by synchronous and asynchronous correlation spectrum changes in 2D-COS was analyzed, and local sensitive variables closely related to Ala content were obtained. On this basis, the simplified linear, nonlinear, and artificial neural network models developed by the weighted coefficient based on the feature wavelength extraction method were compared. The results show that with the change in Ala content in beef, the double-frequency absorption of the C-H bond of CH2 in the chemical bond sequence occurred prior to the third vibration of the C=O bond and the first stretching of O-H in COOH. Furthermore, the wavelength within the 1136-1478 nm spectrum range was obtained as the local study area of Ala content. The linear partial least squares regression (PLSR) model based on effective wavelengths was selected by competitive adaptive reweighted sampling (CARS) from 2D-COS analysis, and provided excellent results (R2C of 0.8141, R2P of 0.8458, and RPDp of 2.54). Finally, the visual distribution of Ala content in beef was produced by the optimal simplified combination model. The results show that 2D-COS combined with NIR-HSI could be used as an effective method to monitor Ala content in beef.

Evaluation of weather parameter-based pre-harvest yield forecast models for wheat crop: a case study in Saurashtra region of Gujarat.

Wheat is the important food grain and is cultivated in many Indian states: Punjab, Haryana, Uttar Pradesh, and Madhya Pradesh, which contributes to major crop production in India. In this study, popular statistical approach multiple linear regression (MLR) and time series approaches Time Delay Neural Network (TDNN) and ARIMAX models were envisaged for wheat yield forecast using weather parameters for a case study area, i.e., Junagarh district, western Gujarat region situated at the foot of Mount Girnar. Weather data corresponds to 19weeks (42nd to 8th Standard Meteorological Week, SMW) during crop growing season was used for prediction of wheat yield using these statistical techniques and were evaluated for their predictive capability. Furthermore, trend analysis among weather parameters and crop yield was also carried out in this study using non-parametric Mann-Kendall test and Sen's slope method. Significant negative correlation was observed between wheat yield and some of the weekly weather variables, viz., maximum temperature (48, 49, 50, 51, 52, and 4th SMW), and total rainfall (50, 51, and 1st SMW) while positive correlation was observed with morning relative humidity (49 and 3rd SMW). Study indicated that forecast error varied from 1.80 to 10.28 in MLR, 0.79 to 7.79 in ARIMAX (2,2,2), - 3.09 to 10.18 in TDNN (4,5) during model training period (1985-2014). The MAPE value shows that the time series data predicted less than 5% of variation, whereas the conventional MLR technique indicated more than 7% variation. Both ARIMAX and TDNN approaches indicated better performance during model training periods, i.e., 1985-2014 and 1985-2015, while former performed well during the forecast periods 1985-2016 and 1985-2017. Overall, the study indicated that the ARIMAX approach can be used consistently for 4years using the same model.

Fast Algorithms for Exact IR Drop De-Embedding in Analog Multiply–Accumulate Computing

IR drop that comes from the line resistance is a well-known issue of the crossbar structure. When the latter is used in the in-memory computing, this issue sets the limit for the analog computing parallelism. In this work, the previously unappreciated linear network modeling is proven to be sufficient for its calibration. Fast algorithms exploiting this simplification are proposed for an exact and efficient de-embedding of the IR drop. Physical pictures are offered for a more intuitive understanding. Scaling in the input and output peripherals is proposed to overcome the dynamic range overflow issue while maintaining the memristor cell precision. An all-in-one-chip MNIST demonstration is presented. Neural network scheduling and gear-like convolution are proposed for real-time processing. Full MNIST test set with the calibration shows a 98.70% accuracy, with 0.45% loss from the software baseline and a 3.58% improvement upon the without-calibration case. Besides inference deployment, these methods may also be generalized to edge learning where the computation resources are limited.

Network model for predicting structural properties of paper

Abstract Paper simulations that resolve the entire microscopic fiber structure are typically time-consuming and require extensive resources. Several such modeling approaches have been proposed to analyze different properties in paper. However, most use non-linear and time-dependent models resulting in high computational complexity. Resolving these computational issues would increase its usefulness in industrial applications. The model proposed in this work was developed in collaboration with companies in the papermaking industry within the Innovative Simulation of Paper (ISOP) project. A linear network model is used for efficiency, where 1-D beams represent the fibers. Similar models have been proposed in the past. However, in this work, the paper models are three-dimensional, a new dynamic bonding technique is used, and more extensive simulations are evaluated. The model is used to simulate tensile stiffness, tensile strength, and bending resistance. These simulated results are compared to experimental and theoretical counterparts and produce representable results for realistic parameters. Moreover, an off-the-shelf computer accessible to a paper developer can evaluate these models structural properties efficiently.

A Hybrid ARFIMA Wavelet Artificial Neural Network Model for DJIA Index Forecasting

This paper proposes a hybrid modelling approach for forecasting returns and volatilities of the stock market. The model, called ARFIMA-WLLWNN model, integrates the advantages of the ARFIMA model, the wavelet decomposition technique (namely, the discrete MODWT with Daubechies least asymmetric wavelet filter) and artificial neural network (namely, the LLWNN neural network). The model develops through a two-phase approach. In phase one, a wavelet decomposition improves the forecasting accuracy of the LLWNN neural network, resulting in the Wavelet Local Linear Wavelet Neural Network (WLLWNN) model. The Back Propagation and Particle Swarm Optimization (PSO) learning algorithms optimize the WLLWNN structure. In phase two, the residuals of an ARFIMA model of the conditional mean become the input to the WLLWNN model. The hybrid ARFIMA-WLLWNN model is evaluated using daily returns of the Dow Jones Industrial Average index over 01/05/2010 to 02/11/2020. The experimental results indicate that the PSO-optimized version of the hybrid ARFIMA-WLLWNN outperforms the LLWNN, WLLWNN, ARFIMA-LLWNN, and the ARFIMA-HYAPARCH models and provides more accurate out-of-sample forecasts over validation horizons of one, five and twenty-two days.

Data-driven strategies for extractive distillation unit optimization

We provide insights for the development of a fast, dynamic, and adaptive way of modeling and optimizing chemical processes. We investigate two data-driven methodologies to optimize the energy cost of an extractive distillation process using an Aspen simulator. The first method uses surrogate-based optimization to explore two model-building techniques: Automated Learning of Algebraic Models’s (ALAMO) generalized linear models and neuron network models with rectifier (ReLU) activation function. We compare the accuracy and performance of these models when embedded into mathematical programming models. The second approach uses black-box optimization (BBO) to optimize problems directly using simulation results. We compare four BBO solvers and three different penalty functions to address constraints. We find that ALAMO performs well for a less complex system, whereas the ReLU network performs better for a more complex one. BBO with a smooth penalty function for constraints is the more effective approach for problems considered in this study.

A two-layer transformation for zeros analysis of networked dynamical systems

Input–output properties are studied for a linear network model with homogeneous single-input single-output subsystems. Specifically, the finite- and infinite- zero structure of a channel in the network, defined by a single external actuation and measurement, is characterized. This is done by developing a two-layer transformation for the input–output model, which expresses the zero structure in terms of the subsystem model, global network interactions (graph), and the input and output locations. This algebraic characterization then provides a means to: (1) distinguish structural properties of the input–output channel (minimum and nonminimum behavior) in terms of the network’s graph, and (2) design of the graph to shape the input–output dynamics. Similar characterizations are also briefly explored for a multiple-input multiple-output (MIMO) network model with collocated measurements and actuations.

Use of geostatistical models to evaluate landscape and stream network controls on post‐fire stream nitrate concentrations

AbstractForested watersheds provide many ecosystem services, such as the filtration of sediment, pollutants, and nutrients, which are increasingly threatened by wildfire. For example, stream nutrient concentrations often increase following wildfire and can remain elevated for decades, making downstream waters susceptible to eutrophication. We investigated the drivers of persistent elevated stream nutrients, specifically nitrate (NO3−), in nine watersheds that were burned 16 years prior by the Hayman fire, Colorado, USA. We evaluated the ability of multiple linear regression and spatial stream network modeling approaches to predict observed concentrations of the biologically active solute NO3− and the conservative solute sodium (Na+) which serves as a partial control. Specifically, we quantified the degree to which landscape and stream network characteristics predict stream solute concentrations. Stream Na+ exhibited strong spatial autocorrelation that was primarily controlled by topography and hydrology. In contrast, stream NO3− had higher spatial variability and was inversely correlated to vegetation cover, measured as mean normalized differenced moisture index (NDMI). Spatially heterogeneous wildfire behaviour left intact forest patches interspersed with high burn severity patches dominated by shrubs and grasses which contributes to the spatial variability in stream NO3− concentrations. Post‐fire vegetation also interacts with watershed structure to influence stream NO3− patterns. For example, severely burned convergent hillslopes in headwaters positions were associated with the highest stream NO3− concentrations due to the high proportional influence of hillslope water in these locations. Our findings suggest that targeted reforestation in severely burned convergent hillslopes in headwater positions may enhance the recovery of stream NO3− concentrations to pre‐fire levels.

A Neural-Inspired Architecture for EEG-Based Auditory Attention Detection

Humans have the ability to focus on one of the sound sources in a noisy scene, which is critical for everyday communication. Auditory attention detection (AAD) seeks to detect selective attention from one’s brain signals. For AAD to be useful in brain–computer interface applications, new approaches with low computational cost, high classification performance, and low latency are required to be developed. In this study, we proposed a novel neural-inspired architecture to mimic the neural computation and coding strategy in the brain for electroencephalography-based AAD. We validated our model through data visualization, and conducted experiments on two publicly available databases. For both KUL and DTU databases, it outperforms both linear and convolutional neural network (CNN) models with consistent improvements from 1 s to 5 s decision windows in terms of detection accuracy. Although the accuracy of the proposed neural-inspired model is inferior to the state-of-the-art spatio-spectral feature (SSF)-CNN model, the computational cost of our model is less than 1% of SSF-CNN’s. Moreover, the neural-inspired decoder is more hardware friendly and energy-efficient due to its biological computing scheme. Overall, the proposed neural-inspired architecture realizes a fast, accurate, and low energy expenditure AAD, which is a big step forward towards practical neuro-steered hearing aids.

Rule and Neural Network-Based Image Segmentation of Mice Vertebrae Images.

BackgroundImage segmentation is a fundamental technique that allows researchers to process images from various sources into individual components for certain applications, such as visual or numerical evaluations. Image segmentation is beneficial when studying medical images for healthcare purposes. However, existing semantic image segmentation models like the U-net are computationally intensive. This work aimed to develop less complicated models that could still accurately segment images.MethodologyRule-based and linear layer neural network models were developed in Mathematica and trained on mouse vertebrae micro-computed tomography scans. These models were tasked with segmenting the cortical shell from the whole bone image. A U-net model was also set up for comparison.ResultsIt was found that the linear layer neural network had comparable accuracy to the U-net model in segmenting the mice vertebrae scans.ConclusionsThis work provides two separate models that allow for automated segmentation of mouse vertebral scans, which could be potentially valuable in applications such as pre-processing the murine vertebral scans for further evaluations of the effect of drug treatment on bone micro-architecture.