Basic Model Research Articles

Image Forgery Detection Based on Parallel Convolutional Neural Networks

Abstract— Due to the availability of deep networks, progress has been made in the field of image recognition. Images and videos are spreading very conveniently and with the availability of strong editing tools the tampering of digital content become easy. To detect such scams, we proposed techniques. In our paper, we proposed two important aspects of employing deep convolutional neural networks to image forgery detection. We first explore and examine different preprocessing method along with convolutional neural networks (CNN) architecture. Later we evaluated the different transfer learning for pre-trained ImageNet(via-fine-tuning) and implement it over our dataset CASIA V2.0. So, it covers the pre-processing techniques with basic CNN model and later see the powerful effect of the transfer learning models. Keywords— image tampering, convolution neural network (CNN), error level analysis (ELA), transfer learning, sharpening filter, fine-tuning

Utility of PBPK Modeling in Predicting and Characterizing Clinical Drug Interactions.

Physiologically-based pharmacokinetic (PBPK) modeling is a mechanistic dynamic modeling approach that can be used to predict or retrospectively describe changes in drug exposure due to drug-drug interactions. With advancements in commercially available PBPK software, PBPK DDI modeling has become a mainstream approach from early drug discovery through to late stage drug development and is often utilized to support regulatory packages for new drug applications. This minireview will briefly describe the approaches to predicting DDI utilizing PBPK and static modeling approaches, the basic model structures and features inherent to PBPK DDI models and key examples where PBPK DDI models have been used to describe complex DDI mechanisms. Future directions aimed at using PBPK models to characterize transporter-mediated DDI, predict DDI in special populations and assess the DDI potential of protein therapeutics will be discussed. A summary of the 209 PBPK DDI examples published to date in 2023 will be provided. Overall, current data and trends suggest a continued role for PBPK models in the characterization and prediction of DDI for therapeutic molecules. Significance Statement PBPK models have been a key tool in the characterization of various pharmacokinetic phenomenon, including drug-drug interactions. This minireview will highlight recent advancements and publications around PBPK DDI modeling, an important area of drug discovery and development research in light of the increasing prevalence of polypharmacology in clinical settings.

Improving risk models for patients having emergency bowel cancer surgery using linked electronic health records: a national cohort study.

Life-saving emergency major resection of colorectal cancer (CRC) is a high-risk procedure. Accurate prediction of postoperative mortality for patients undergoing this procedure is essential for both healthcare performance monitoring and preoperative risk assessment. Risk-adjustment models for CRC patients often include patient and tumour characteristics, widely available in cancer registries and audits. The authors investigated to what extent inclusion of additional physiological and surgical measures, available through linkage or additional data collection, improves accuracy of risk models. Linked, routinely-collected data on patients undergoing emergency CRC surgery in England between December 2016 and November 2019 were used to develop a risk model for 90-day mortality. Backwards selection identified a 'selected model' of physiological and surgical measures in addition to patient and tumour characteristics. Model performance was assessed compared to a 'basic model' including only patient and tumour characteristics. Missing data was multiply imputed. Eight hundred forty-six of 10578 (8.0%) patients died within 90 days of surgery. The selected model included seven preoperative physiological and surgical measures (pulse rate, systolic blood pressure, breathlessness, sodium, urea, albumin, and predicted peritoneal soiling), in addition to the 10 patient and tumour characteristics in the basic model (calendar year of surgery, age, sex, ASA grade, TNM T stage, TNM N stage, TNM M stage, cancer site, number of comorbidities, and emergency admission). The selected model had considerably better discrimination compared to the basic model (C-statistic: 0.824 versus 0.783, respectively). Linkage of disease-specific and treatment-specific datasets allowed the inclusion of physiological and surgical measures in a risk model alongside patient and tumour characteristics, which improves the accuracy of the prediction of the mortality risk for CRC patients having emergency surgery. This improvement will allow more accurate performance monitoring of healthcare providers and enhance clinical care planning.

New nonlinear models of dynamic longitudinal deformation of a viscoelastic rod

In this paper we study a general model of dynamic longitudinal deformation of a viscoelastic rod with a nonlinear dependence of the stress on the strain and on the strain rate. This model is given by a nonlinear differential equation with partial derivatives of the third order. A group classification of this model which depends on two arbitrary functions is carried out. Six basic models are obtained, for which their main Lie groups of transformations are found. A model that admits the widest Lie group of transformations is studied in detail. For it, formula for the production of new solutions is obtained and all invariant submodels are found. For all its invariant submodels, the invariant solutions defining these submodels are found either in explicit form, or their search reduces to solving the systems of first-order differential equations. For these systems, boundary value problems that have a physical meaning are studied. Conditions are obtained that ensure the existence and uniqueness of solutions to these boundary value problems. This allows one to correctly solve these boundary value problems numerically. Boundary value problems for some specific values of the parameters included in them are solved numerically. The research carried out is especially relevant in rocket engineering, aircraft engineering, shipbuilding and other fields.

Kernel‐PCA‐based single‐phase earth fault detection model using multilayer perceptron in deep learning

AbstractThe rapid and accurate identification of single‐phase earth fault is of prime concern for many scholars from industry and academia, because its fault characteristics are extremely weak while the related potential hazard is extremely severe. In this context, this paper borrows from Deep Learning and in turn innovatively puts forward a multilayer perceptron (MLP)‐based single‐phase earth fault detection model augmented with kernel principal component analysis (KPCA). First, KPCA is applied to building fault feature extraction engineering via the transformation of fault features from low‐dimensional linear indivisible space to high‐dimensional one. Second, a MLP‐based model is used for basic model reference where many optimization strategies together with dropout technique are introduced and custom‐designed in tuning this model to achieve the best detection accuracy and convergence characteristic. Numerical studies demonstrate the superiority of the proposed KPCA‐MLP‐based single‐phase earth fault detection model.

ML-Driven Facial Synthesis from Spoken Words Using Conditional GANs

A Human Brain may translate a person's voice to its corresponding face image even if never seen before. Training adeep learning network to do the same can be used in detecting human faces based on their voice, which may be used in findinga criminal that we only have a voice recording for. The goal in this paper is to build a Conditional Generative Adversarial Network that produces face images from human speeches which can then be recognized by a face recognition model to identifythe owner of the speech. The model was trained, and the face recognition model gave an accuracy of 80.08% in training and 56.2% in testing. Compared to the basic GAN model, this model has improved the results by about 30%. Key Word: Face image synthesis, Generative adversarial network, Face Recognition

Property parameter determination in individual layers for separately fractured wells with commingled production in multi-layered reservoirs

At present, without any separate rate test for each layer, there is no way to determinate the properties of individual layers for separately fractured wells with commingled production in multi-layered reservoirs. In order to address this issue, much research work was performed and elucidated in this article. To begin with, we illustrated a basic physical model for a separately fractured well in a multi-layered reservoir. Next, we stated the common determination method that can only be used to gain the average properties of multi-layered reservoirs. Then, according to the physical model, we newly established a mathematical model and plotted standard well-test type curves; additionally, we also discussed why we cannot determinate the properties of individual layers by using the new well-test model. What’s more, we presented a new method to determinate the properties of individual layers. Moreover, we compared the advantages and disadvantages among the three methods. In addition, by using the new determination method, we particularly took two field wells as examples to demonstrate how to determine the properties of individual layers. The proposed new method was validated by use of the common method, the new well-test model and the microseismic monitoring results. At the end, we summarized the research conclusions and indicated that the new method was a good tool to determinate the properties of individual layers in multi-layered reservoirs.

An integrated GBWM-PROMETHEE-CLOUD & MCGP model for green supplier selection and order allocation (GSSOA) in an oil refinery

With the increase of environmental concerns in recent years and the prevalence of such issues, which originate mainly from large industries and manufacturing plants, green supply chain management (GSCM) issues have been given special attention among researchers. GSCM is a group of management applications integrating supply chain management with criteria related to environment and supplier selection with the purpose of reducing environmental pollution caused by processes involved in a company supply chain. In this research the main contributions is to draw a decision-making structure addressing the applicability of the green supplier selection (GSS) and order allocation (OA) problem in an Iranian Oil refinery company. A combination of Cloud-PROMETHEE along with using GBWM was taken into account for GSS as the first phase and MCGP method for OA as the second phase, in order to solve the GSSOA problem. The proposed technique lets the vagueness and uncertainty of experts' judgments be obtained by assisting uncertain linguistic variable derived from the cloud model. Furthermore, the group best-worst method (GBWM) obtains the optimal criteria weights by forming a nonlinear programing model and the prioritizing of nominated suppliers is established using the PROMETHEE method with the possibility to pick a distinct preference function for each criterion and draw a graph of supplier preferences. At last, MCGP method was applied considering with the possibility of multi-choice aspiration levels (MCAL) for each goal. The suggested model illustrated with actual figures in an oil refinery to provide insight to decision-makers of green supplier selection and order allocation. To do so, initially nine criteria have been derived after vast research on literature and experts’ opinions. Then the suppliers was successfully evaluated with considering uncertainties and randomness in a group of decision maker judgments. The order allocation was also completed by minimizing total cost and delays and maximizing purchase using results of supplier ranking in the first phase. Results have shown that the supplier with better ranking in GSS is preferred in allocating orders. A sensitivity analysis then performed and confirmed that the suggested method is highly stable against different changes in the data. The proposed method was also compared with the basic MCGDM cloud model and the method is approved. Finally, managerial implications and conclusions for additional researches are demonstrated.

A significance score for protein-protein interaction models through random docking.

Comparing accuracies of structural protein-protein interaction (PPI) models for different complexes on an absolute scale is a challenge, requiring normalization of scores across structures of different sizes and shapes. To help address this challenge, we have developed a statistical significance metric for docking models, called random-docking (RD) p-value. This score evaluates a PPI model based on how likely a random docking process is to produce a model of better or equal accuracy. The binding partners are randomly docked against each other a large number of times, and the probability of sampling a model of equal or greater accuracy from this reference distribution is the RD p-value. Using a subset of top predicted models from CAPRI (Critical Assessment of PRediction of Interactions) rounds over 2017-2020, we find that the ease of achieving a given root mean squared deviation or DOCKQ score varies considerably by target; achieving the same relative metric can be thousands of times easier for one complex compared to another. In contrast, RD p-values inherently normalize scores for models of different complexes, making them globally comparable. Furthermore, one can calculate RD p-values after generating a reference distribution that accounts for prior information about the interface geometry, such as residues involved in binding, by giving the random-docking process access the same information. Thus, one can decouple improvements in prediction accuracy that arise solely from basic modeling constraints from those due to the rest of the method. We provide efficient code for computing RD p-values at https://github.com/Grigoryanlab/RDP.

Active Learning for Stacking and AdaBoost-Related Models

Ensemble learning (EL) has become an essential technique in machine learning that can significantly enhance the predictive performance of basic models, but it also comes with an increased cost of computation. The primary goal of the proposed approach is to present a general integrative framework that allows for applying active learning (AL) which makes use of only limited budget by selecting optimal instances to achieve comparable predictive performance within the context of ensemble learning. The proposed framework is based on two distinct approaches: (i) AL is implemented following a full scale EL, which we call the ensemble learning on top of active learning (ELTAL), and (ii) apply the AL while using the EL, which we call the active learning during ensemble learning (ALDEL). Various algorithms for ELTAL and ALDEL are presented using Stacking and Boosting with various algorithm-specific query strategies. The proposed active learning algorithms are numerically illustrated with the Support Vector Machine (SVM) model using simulated data and two real-world applications, evaluating their accuracy when only a small number instances are selected as compared to using full data. Our findings demonstrate that: (i) the accuracy of a boosting or stacking model, using the same uncertainty sampling, is higher than that of the SVM model, highlighting the strength of EL; (ii) AL can enable the stacking model to achieve comparable accuracy to the SVM model using the full dataset, with only a small fraction of carefully selected instances, illustrating the strength of active learning.

Common infections and neuroimaging markers of dementia in three UK cohort studies.

We aimed to investigate associations between common infections and neuroimaging markers of dementia risk (brain volume, hippocampal volume, white matter lesions) across three population-based studies. We tested associations between serology measures (pathogen serostatus, cumulative burden, continuous antibody responses) and outcomes using linear regression, including adjustments for total intracranial volume and scanner/clinic information (basic model), age, sex, ethnicity, education, socioeconomic position, alcohol, body mass index, and smoking (fully adjusted model). Interactions between serology measures and apolipoprotein E (APOE) genotype were tested. Findings were meta-analyzed across cohorts (Nmain =2632; NAPOE-interaction =1810). Seropositivity to John Cunningham virus associated with smaller brain volumes in basic models (β=-3.89mL [-5.81, -1.97], Padjusted <0.05); these were largely attenuated in fully adjusted models (β=-1.59mL [-3.55, 0.36], P=0.11). No other relationships were robust to multiple testing corrections and sensitivity analyses, but several suggestive associations were observed. We did not find clear evidence for relationships between common infections and markers of dementia risk. Some suggestive findings warrant testing for replication.

Cracking the black box of deep sequence-based protein-protein interaction prediction.

Identifying protein-protein interactions (PPIs) is crucial for deciphering biological pathways. Numerous prediction methods have been developed as cheap alternatives to biological experiments, reporting surprisingly high accuracy estimates. We systematically investigated how much reproducible deep learning models depend on data leakage, sequence similarities and node degree information, and compared them with basic machine learning models. We found that overlaps between training and test sets resulting from random splitting lead to strongly overestimated performances. In this setting, models learn solely from sequence similarities and node degrees. When data leakage is avoided by minimizing sequence similarities between training and test set, performances become random. Moreover, baseline models directly leveraging sequence similarity and network topology show good performances at a fraction of the computational cost. Thus, we advocate that any improvements should be reported relative to baseline methods in the future. Our findings suggest that predicting PPIs remains an unsolved task for proteins showing little sequence similarity to previously studied proteins, highlighting that further experimental research into the 'dark' protein interactome and better computational methods are needed.

Full elucidation of sorting mechanisms in and around the Golgi apparatus by super-resolution live imaging

The Golgi body or Golgi apparatus is a cell organelle that plays a key role in protein transport and sorting in the cell. Although basic models of ‘vesicular transport’ processes appear to be widely accepted, there are many researchers who question the status quo. Because of the difficulties of working with traffic processes in living cells, there are still many unanswered questions. Dr Akihiko Nakano is a researcher working to uncover the full truth about protein transport within the Golgi apparatus. Indeed, his work on the Golgi cisternal maturation is viewed as a monumental milestone in the field. Technological developments, particularly in live imaging microscopy, mean it is possible to look at specific objects in unprecedented detail and these advancements could enable Nakano to fully elucidate the sorting mechanisms in and around the Golgi apparatus. Nakano is based at the RIKEN Center for Advanced Photonics (RAP) in Japan and forms part of the Live Cell Super-Resolution Imaging Research Team. He and the team are harnessing the power and potential of live imaging to re-examine the processes of membrane traffic in living cells of different species, yeast, plant and animal cells. The goal is to propose a fundamental model of membrane trafficking that can explain seemingly different behaviours of membranes by common mechanisms. A key part of the teamâ–™s research has involved the development of a novel microscopic methodology called super-resolution confocal live imaging microscopy (SCLIM).

Coupling effect of non‐relaxing and high temperature on fatigue properties of carbon‐black filled isoprene rubber

AbstractThis paper presents an extensive experimental investigation on fatigue properties of carbon‐black filled isoprene rubber under complex loadings. A basic life prediction model taking strain energy density as fatigue parameter is first proposed based on fatigue crack growth tests and uniaxial tension fatigue tests under relaxing loads at room temperature. A database of fatigue life including relaxing and non‐relaxing is established through a great number of fatigue tests under five temperatures. Based on the database, two empirical parameters, that is, temperature factor (NRT/NHT) and life reinforcement factor (Re), are introduced to quantitatively characterize the coupling effects of high temperature and non‐relaxing loads on fatigue life. The fatigue mechanisms under different conditions are compared via wide‐angle x‐ray diffraction tests and postmortem analysis. It reveals that the weakening of life reinforcement at high temperatures is closely related to the strain induced crystallization behavior of rubber.

Electric Submersible Pump Lifted Oil Field: Basic Model for Control, and Comparison of Simulation Tools

Optimal operation of petroleum production is important in a transition from energy systems based on fossil fuel to sustainable systems. One sub-process in petroleum production deals with transport from the (subsea) well-bore to a topside separator. Good control design for such operation requires a dynamic model of the petroleum flow from the well-bore to the separator. Here, such a dynamic model is considered for liquid production (oil/water) using an electric submersible pump (ESP) to aid in counteracting gravity and friction forces. Based on an existing model used for industrial control design, one goal is to report a complete dynamic model in a single paper. Emphasis is put on dimensionless equipment models for the simple change of units, and the model is developed from physical laws for easy replacement of sub-models, if needed. All the necessary information (equations, parameters) for model implementation is provided, and two candidate equation-based modeling languages are selected and compared: Modelica and ModelingToolkit [MTK] for Julia. The simulation results are virtually identical for the two languages and make sense from physics; however, there is a minor discrepancy in one plot—likely caused by slight differences in accuracy in handling initialization in the implicit algebraic equations. The implementation structures of the model in Modelica and MTK are similar. Modelica is a mature and excellent modeling tool, handles large-scale models, and has tools for producing C code and integration with other tools. MTK is still in rapid development, supports more model types than Modelica, and is integrated in an eco-system with excellent support for control design, optimization, model fitting, and more. To illustrate the suitability of using the developed model for control design, a simple PI controller is designed within the eco-system of MTK/Julia.

Cross-channel influences in mobile-app-website e-commerce supply chains: When to weaken the influence?

Today, e-commerce has entered the mobile era in which consumers can shop through both mobile app (MA) and website (WS) channels. By strategically designing the linkage between the two channels based on the cross-channel influences, companies can better handle the demand uncertainty of seasonal products. In this paper, we consider a dual channel e-commerce supply chain with an e-tailer and a manufacturer selling a seasonal product. We analytically examine how the e-tailer can (i) adopt risk pooling by aggregating the demands from both the MA and WS channels and (ii) invest in forecast-enhancement technology (FET) to improve inventory management. We uncover that when the magnitude of cross-channel influences increases, the impacts on the optimal inventory decision vary greatly depending on four different “model cases”; interestingly, it does not affect the optimal FET decision. Considering the impacts of cross-channel influences on the e-tailer’s expected profit, we reveal that the optimal weakening or strengthening strategy, i.e., “creating barriers” or “enhancing linkage” between two channels, depends on the e-tailer’s channel operations models. Then, we find that the optimal weakening or strengthening strategy follows the same pattern as in the basic model when the dual channel e-commerce supply chain is coordinated under a Nash bargaining framework. Last but not least, we study the potential use of blockchain technology to improve the effectiveness of FET investment. We surprisingly find that using blockchain or not does not affect the optimal MA and WS design decisions on weakening or strengthening linkages.

A modified high C-rate battery equivalent circuit model based on current dependence and concentration modification

The current battery equivalent circuit models are mainly suitable for room temperature or low C-rate conditions. However, the polarization phenomenon is serious at high C-rate, which results in the obvious voltage drop and significantly reduces the model accuracy at high C-rate. In this paper, the characteristics analysis of different RC models is firstly studied. It is found that the first-order RC circuit model is more suitable for high C-rate conditions due to the coupling of model parameter extraction. Based on the first-order RC model, a basic equivalent circuit model adapted to high C-rate condition is then determined. Secondly, the influences of sampling frequency and the identified time domain on instantaneous impedance and polarization resistance are researched. Thirdly, the change rules of different polarizations at high C-rate are analyzed. Furthermore, the relationship between the change rule of the model parameters and that of different overpotentials are researched. Subsequently, a modified high C-rate model is innovatively established, which considers the current dependence and concentration/temperature modification of instantaneous impedance and polarization resistance. Finally, the accuracy of the modified high C-rate model is verified at 5 °C, 25 °C, and 55 °C, respectively. Results show that the modified high C-rate model can improve the simulation accuracy of the voltage at high C-rate (≤6C), and the mean absolute error (MAE) and Root Mean Squared Error (RMSE) are within 68 mV. Compared with the existing high C-rate model, the accuracy of the modified high C-rate model proposed in this paper is improved and the complexity is also significantly reduced.

A gentle introduction to computer vision-based specimen classification in ecological datasets.

Classifying specimens is a critical component of ecological research, biodiversity monitoring and conservation. However, manual classification can be prohibitively time-consuming and expensive, limiting how much data a project can afford to process. Computer vision, a form of machine learning, can help overcome these problems by rapidly, automatically and accurately classifying images of specimens. Given the diversity of animal species and contexts in which images are captured, there is no universal classifier for all species and use cases. As such, ecologists often need to train their own models. While numerous software programs exist to support this process, ecologists need a fundamental understanding of how computer vision works to select appropriate model workflows based on their specific use case, data types, computing resources and desired performance capabilities. Ecologists may also face characteristic quirks of ecological datasets, such as long-tail distributions, 'unknown' species, similarity between species and polymorphism within species, which impact the efficacy of computer vision. Despite growing interest in computer vision for ecology, there are few resources available to help ecologists face the challenges they are likely to encounter. Here, we present a gentle introduction for species classification using computer vision. In this manuscript and associated GitHub repository, we demonstrate how to prepare training data, basic model training procedures, and methods for model evaluation and selection. Throughout, we explore specific considerations ecologists should make when training classification models, such as data domains, feature extractors and class imbalances. With these basics, ecologists can adjust their workflows to achieve research goals and/or account for uncertainty in downstream analysis. Our goal is to provide guidance for ecologists for getting started in or improving their use of machine learning for visual classification tasks.

A Hybrid Model for Forecasting Realized Volatility Based on Heterogeneous Autoregressive Model and Support Vector Regression

In this study, we proposed two types of hybrid models based on the heterogeneous autoregressive (HAR) model and support vector regression (SVR) model to forecast realized volatility (RV). The first model is a residual-type model, where the RV is first predicted using the HAR model, and the residuals are used to train the SVR model. The residual component is then predicted using the SVR model, and the results from both the HAR and SVR models are combined to obtain the final prediction. The second model is a weight-based model, which is a combination of the HAR and SVR models and uses the same independent variables and dependent variables as the HAR model; we adjust the contribution of the two models to the predicted values by giving different weights to each model. In particular, four volatility models are used in RV forecasting as basic models. For empirical analysis, the RV of returns of the Tokyo stock price index and five individual stocks of TOPIX 30 is used as the dataset. The empirical results reveal that according to the model confidence set test, the weight-type model outperforms the HAR model and the residual-type HAR–SVR model.

EFFECT OF CHANGES IN THE LENGTH OF THE COMPONENTS OF THE MUSCULOTENDINOUS ELEMENT OF THE ELBOW FLEXOR MUSCLES ON THE ISOMETRIC FORCE AND JOINT TORQUE

Immobilization of the joint leads to the formation of immobilization contracture, which is accompanied by a decrease in the elasticity of tendons and muscles, i.e. loss of full contraction and stretching. The torque in human joints is one of the key indicators in assessing rehabilitation. Objective. To study the effect of changes in the strength, length of muscles and tendons of the elbow joint on the torque in flexion. Methods. The basic OpenSim model arm26 was used for modeling. To determine the change in the length of the components of the muscle-tendon element (MTE), their length was determined at a 90° angle of elbow flexion. The decrease in muscle strength was considered a loss per day for elbow flexors — 1.2 %, extensors — 1.1 %. The decrease in strength was calculated for a period of immobilization of 45 days. Three models were created: Normal — a model without changes in muscle parameters; Contracture — a change in the length of muscles and tendons; Contracture + muscle (CM) — an additional decrease in muscle strength. Results. The obtained data of torques when changing the length of the MTE components showed their increase in conditions of unchanged isometric muscle strength. But this option is not possible after immobilization of the limb. Therefore, it is closer to the real model of СM, in which the torque is significantly reduced by the amount of decrease in muscle strength. These models show a tendency that the change in the components of the MTE due to immobilization increases the joint torque and, when trying to apply excessive force during joint development, can lead to traumatic consequences. During immobilization, the flexor muscles shorten, which prevents the patient from fully extending the elbow joint. Conclusions. This work on predicting the elbow joint torque generated by the muscles can be useful in studying specific clinical situations with elbow joint contractures, but cannot be fully transferred to practice due to the significant conventionality of the model parameters. However, the modeling method can show trends in changes in muscle function parameters when their geometry changes.