Network Model Research Articles

Learning and generating vehicle motion primitives from human-driving data

Abstract Motion primitives play an important role in motion planning for autonomous vehicles, as they effectively address the sampling challenges inherent in nonholonomic motion planning. Employing motion primitives (MPs) is a widely accepted approach in nonholonomic motion planning based on sampling. This study specifically addresses the problem of learning from human-driving data to create human-like trajectories from predefined start-to-end states, which then serve as MP within the sampling-based nonholonomic motion planning framework. In this paper, we propose a deep learning-based method for generating MP that capture human-driving trajectory data features. By processing human-driving trajectory data, we create a Motion Primitive dataset that uniformly covers typical urban driving scenarios. Based on this dataset, a vehicle model long short-term memory neural network model is constructed to learn the features of the human-driving trajectory data. Finally, a framework for the generation of MP for practical applications is given based on this neural network. Our experiments, which focus on the dataset, the MMP generation network, and the generation process, demonstrate that our method significantly improves the training efficacy of the MP generation network. Additionally, the MP generated by our method exhibit higher accuracy compared to traditional methods.

Accurate Neural Network Fine-Tuning Approach for Transferable Ab Initio Energy Prediction across Varying Molecular and Crystalline Scales.

Existing machine learning models attempt to predict the energies of large molecules by training small molecules, but eventually fail to retain high accuracy as the errors increase with system size. Through an orbital pairwise decomposition of the correlation energy, a pretrained neural network model on hundred-scale data containing small molecules is demonstrated to be sufficiently transferable for accurately predicting large systems, including molecules and crystals. Our model introduces a residual connection to explicitly learn the pairwise energy corrections, and employs various low-rank retraining techniques to modestly adjust the learned network parameters. We demonstrate that with as few as only one larger molecule retraining the base model originally trained on only small molecules of (H2O)6, the MP2 correlation energy of the large liquid water (H2O)64 in a periodic supercell can be predicted at chemical accuracy. Similar performance is observed for large protonated clusters and periodic poly glycine chains. A demonstrative application is presented to predict the energy ordering of symmetrically inequivalent sublattices for distinct hydrogen orientations in the ice XV phase. Our work represents an important step forward in the quest for cost-effective, highly accurate and transferable neural network models in quantum chemistry, bridging the electronic structure patterns between small and large systems.

Breast cancer image classification based on H&E staining using a causal attention graph neural network model

Breast cancer image classification based on H&E staining using a causal attention graph neural network model

Photovoltaic solar energy prediction using the seasonal-trend decomposition layer and ASOA optimized LSTM neural network model

As the global energy demand continues to produce, photovoltaic (PV) solar energy has emerged as a key Renewable Energy Source (RES) due to its sustainability and potential to reduce dependence on fossil fuels. However, accurate forecasting of Solar Energy (SE) output remains a significant challenge due to the inherent variability and intermittency of solar irradiance (SI), which is affected by factors such as weather conditions, geographic location, and seasonal patterns. Reliable prediction models are crucial for optimizing energy management, ensuring grid stability, and minimizing operational costs. To address these challenges, this research introduces an innovative method that integrates Robust Seasonal-Trend Decomposition (RSTL) with an Adaptive Seagull Optimisation Algorithm (ASOA)-optimized Long Short-Term Memory (LSTM) neural network. Using RSTL to differentiate between time series data into development, seasonal in nature, and residual factors, this methodology addresses SI’s unpredictable nature and intermittent operation and provides the basis for accurate predictions. ASOA improves LSTM features by constantly finding and exploiting resources and adopting motivation from seagulls’ collecting and migration behaviours. Parameter standardization employing ASOA, the RSTL decomposition approach, and the conceptual model of LSTM networks are all presented in this research work. The proposed method has been contrasted with conventional methods by applying a testing environment incorporating essential Meteorological Factors (MF) and historical SE datasets. The study of performance measurements (RMSE, MAE, and R2) demonstrates significant improvements in the accuracy of predictions. The research results highlight significant implications regarding subsequent studies and real-world uses in SE prediction, accentuating the positive impacts of incorporating accurate data decomposition and adaptive optimized performance.

Intelligent deep federated learning model for enhancing security in internet of things enabled edge computing environment

In the present scenario, the Internet of Things (IoT) and edge computing technologies have been developing rapidly, foremost to the development of new tasks in security and privacy. Personal information and privacy leakage have become the main concerns in IoT edge computing surroundings. The promptly developing IoT-connected devices below an integrated Machine Learning (ML) method might threaten data confidentiality. The standard centralized ML-assisted methods have been challenging because they require vast numbers of data in a vital unit. Due to the rising distribution of information in many systems of linked devices, decentralized ML solutions have been required. Federated learning (FL) was proposed as an optimal solution to discover these privacy issues. Still, the heterogeneity of systems in IoT edge computing environments poses an essential task when executing FL. Therefore, this paper develops an Intelligent Deep Federated Learning Model for Enhancing Security (IDFLM-ES) approach in the IoT-enabled edge-computing environment. The presented IDFLM-ES approach aims to identify unwanted intrusions to certify the safety of the IoT environment. To accomplish this, the IDFLM-ES technique introduces a federated hybrid deep belief network (FHDBN) model using FL on time series data produced by the IoT edge devices. Besides, the IDFLM-ES technique uses data normalization and golden jackal optimization (GJO) based feature selection as a pre-processing step. Besides, the IDFLM-ES technique learns the individual and distributed feature representation over distributed databases to enhance model convergence for quick learning. Finally, the dung beetle optimizer (DBO) model is utilized to choose the effectual hyperparameter of the FHDBN model. The simulation value of the IDFLM-ES methodology is verified using a benchmark database. The experimental validation of the IDFLM-ES methodology portrayed a superior accuracy value of 98.24% compared to other models.

Study on disaster-causing probability evaluation of gas pipeline in karst area.

In this paper, a disaster-causing probability evaluation method of gas pipelines in karst areas based on disaster system theory and vulnerability theory is proposed. The hazard evaluation index system of gas pipeline disaster events in karst areas is established from three aspects: the activity of disaster-prone environment, the risk of disaster factors and the vulnerability of disaster-bearing bodies. Combined with the advantages of information transmission and updating of the Bayesian network model, the hazard probability of disaster events is gradually calculated from a multi-level perspective. Based on the structural reliability theory, the vulnerability evaluation function of gas pipelines in karst areas is established by considering the interaction between disaster hazard intensity and disaster resistance ability. Meanwhile, the design checkpoint method and finite element simulation are used to calculate the vulnerability probability. Finally, a new disaster-causing probability evaluation approach for gas pipelines in karst areas is formed, which comprehensively considers the event hazard and the pipeline vulnerability. The work presented in this paper can provide a reference for the safety management and accident prevention of gas pipelines crossing the karst areas.

Ketamine effects on resting state functional brain connectivity in major depressive disorder patients: a hypothesis-driven analysis based on a network model of depression

IntroductionKetamine demonstrates robust and rapidly occurring antidepressant effects in patients with difficult-to-treat major depressive disorder. Ketamine’s antidepressant effects and its impact on functional networks in non-resistant forms of major depressive disorder are expected to provide valuable insight into ketamine’s mechanism of action related to depression.MethodsThis study employs an existing network model of major depressive disorder to investigate the effects of ketamine on resting state connectivity in a therapy-non-resistant major depressive disorder population. In a randomized, double-blind, placebo-controlled, cross-over study, 0.5 mg/kg racemic ketamine or 0.9%NaCl was administered intravenously in 16 MDD patients. We applied resting-state functional magnetic resonance imaging (rs-fMRI) to explore changes in functional brain connectivity directly at 50, 80 and 165 min (acute) and 24 h (delayed) following ketamine administration. A clinician-rated 10-item scale (MADRS) was administered at 165 min and 24 h after ketamine administration. Connections-of-interest (COIs) were based on the previously published corticolimbic-insular-striatalpallidal-thalamic (CLIPST) circuitry model of major depressive disorder.ResultsCompared with placebo, ketamine significantly (p < 0.0014) reduced the mean (SD) MADRS total score from 21.2 (5.9) pre-dose to 10.3 (4.6) 24 h post-dose. At both acute (p < 0.0172) and delayed (p < 0.0488) time points, significant rs-fMRI connectivity changes occurred only in MDD-related COIs as proposed by the CLIPST model. No changes in functional connectivity were found in non-CLIPST connections.DiscussionThis study demonstrates that ketamine specifically affects depression-related circuitry. Analyzing functional connectivity based on a neurocircuitry model of a specific CNS disease and drug action may be an effective approach that could result in a more targeted analysis in future pharmaco-fMRI studies in CNS drug development.

Explainable Artificial Intelligence twin for metaheuristic optimization: double-skin facade with energy storage in buildings

Abstract The double facade solar chimney along with energy storage components is a sustainable building technology that harnesses the power of the sun to regulate indoor temperatures. Extensive research has been conducted on the theoretical simulation of such systems. The novelty of this work is to explore the potential of explainable artificial intelligence in improving the design and optimization of the double-skin solar chimney. The need for this research is owing to the high computational limitations of the physical model of such systems, thus the application of explainable artificial intelligence based upon Artificial Neural Network can address this research gap. The paper solved a validated physical model and demonstrates the suitability of Artificial Neural Network twins as computational-efficient subrogates that can be later used by a multi-objective optimization function to find the optimal design values for the facades of double-skin buildings. The results of the comparison between the physical and the Artificial Neural Network model show the practical advantage of utilizing the digital twin model without compromising accuracy. The results have indicated that the Artificial Neural Network can achieve a high coefficient of determination ranging from 0.921 to 0.999 on different performance indicators which implies a high goodness of fit. Accordingly, the optimization study based upon a non-sorting genetic algorithm (NSGA-II) has indicated a high ventilation rate of 2.86 1/h and an efficiency of 37.21%. The insights of this work have reflected that exploring the societal implications of sustainable building technologies such as the double facade solar chimney through educational initiatives can cultivate a new generation of society who not only understand the technical aspects but also appreciate the broader social and environmental contexts of their work, eventually to have future buildings with integrated passive systems.

Mapping and modeling age-related changes in intrinsic neural timescales

Intrinsic timescales of brain regions exhibit heterogeneity, escalating with hierarchical levels, and are crucial for the temporal integration of external stimuli. Aging, often associated with cognitive decline, involves progressive neuronal and synaptic loss, reshaping brain structure and dynamics. However, the impact of these structural changes on temporal coding in the aging brain remains unclear. We mapped intrinsic timescales and gray matter volume (GMV) using magnetic resonance imaging (MRI) in young and elderly adults. We found shorter intrinsic timescales across multiple large-scale functional networks in the elderly cohort, and a significant positive association between intrinsic timescales and GMV. Additionally, age-related decline in performance on visual discrimination tasks was linked to a reduction in intrinsic timescales in the cuneus. To explain these age-related shifts, we developed an age-dependent spiking neuron network model. In younger subjects, brain regions were near a critical branching regime, while regions in elderly subjects had fewer neurons and synapses, pushing the dynamics toward a subcritical regime. The model accurately reproduced the empirical results, showing longer intrinsic timescales in young adults due to critical slowing down. Our findings reveal how age-related structural brain changes may drive alterations in brain dynamics, offering testable predictions and informing possible interventions targeting cognitive decline.

Prediction of Shear Strength in Anisotropic Structural Planes Considering Size Effects

It is essential to elucidate the shear mechanical behavior of structural planes to assess the risk to rock masses and protect them from shear failure. Current research on shear mechanical behavior is focused on isotropic structural planes with the same lithology on both sides. However, anisotropic structural planes, commonly found in nature, may exhibit unique mechanical behavior that differs from isotropic structural planes. Therefore, it is necessary to study the factors affecting the shear strength of the anisotropic structural planes. In this paper, the direct shear numerical tests on anisotropic structural planes were carried out using the three-dimensional distinct element code (3DEC) based on the laboratory test. The numerical test results illustrate that the error between the peak shear strength of the numerical test and the laboratory test is basically within 10%. The shear stress-displacement curves of the numerical and laboratory tests are similar, which verifies the accuracy of the numerical test. According to the Barton standard sections, anisotropic structural plane models with different roughness and size were established, and the direct shear numerical tests with different normal stresses were carried out. To predict the peak shear strength of the anisotropic structural planes, one hundred and eighty-one sets of direct shear numerical test data were selected. Normal stress, roughness, compressive strength of soft and hard rock masses, basic friction angle of soft and hard rock masses, and structural plane size were used as input parameters to establish a back propagation (BP) neural network model. The research results show that, under identical conditions, the shear strength of the anisotropic structural planes decreases as the structural plane size increases. On the contrary, the shear strength increases with the increasing structural plane roughness and normal stress. For the BP neural network prediction model, the root mean square error (RMSE) and coefficient of determination (R2) of the training set are 0.441 and 0.957. For the test set, the RMSE is 0.489, and R2 is 0.947, which indicates that the predicted values are in good agreement with the actual values.

Концептуальная структура системы поддержки принятия решений ситуационной видеоаналитики для многообъектного распознавания

Information is provided on the growing interest in situational video analytics (SVA) technologies in various fields of activity. The paper describes a trend towards the implementation of SVA systems based on cognitive artificial intelligence technologies, which make it possible to detect an object and its location in a video stream in real time with a high degree of accuracy. The architecture of a decision support system (DSS) with the possibility of multi-object recognition is proposed. The key logical components of the DSS in the field of SVA are highlighted, as well as the functions and purpose of each component are described. The role of the decision maker for implementing multi-object recognition when searching for an object in a video stream is particularly emphasized. The classification of neural networks by types and fields of application is given, and it is revealed that convolutional neural networks are used to solve video analytics problems. Examples of using the YOLOv5 convolutional model in the DSS model database management system in IAS tasks for detecting the presence of objects on video data are considered. A block diagram of the SVA object recognition algorithm has been developed. A series of experiments has been conducted to train a convolutional neural network model on a unique dataset, on an expanded dataset, and with a new additional object. This organization of experiments is aimed at improving the quality and accuracy of object recognition and exploring the possibility of multi-object recognition. As a result of the experiment, the final trained neural network model was tested and its potential capabilities were analyzed for use in the SVA DSS, taking into account the accuracy of the model. The accuracy was 91.6% for the validation set containing 2 objects. The results obtained using a trained neural network of the YOLOv5 architecture confirms the importance of convolutional neural networks as a key component of the database of SVA DSS models.

Тестирование эффективности гибридной методики шумоподавления в речевом сигнале для системы видеоконференций

The issue of audio signal quality during video conferences is considered. The effect of noise on the quality and intelligibility of the speech signal is described. The analysis of the noise reduction process in the audio signal in real time is carried out. The main approaches to solving the problem of noise reduction in real time are highlighted, namely the approach with recognition and elimination of noise and the approach with voice recognition and elimination of sounds different from the speech signal. An algorithm for intelligent noise reduction using artificial intelligence methods has been developed, characterized by the presence of mechanisms for adjusting the “intensity” of the noise reduction effect on the speech signal, as well as adjusting the “sensitivity” to noise, allowing to influence the efficiency of the algorithm for specific operating conditions without the need to retrain the neural network model of the algorithm. Software has been developed that implements a hybrid noise reduction technique in the form of a module that was implemented in a previously developed software package for organizing and conducting video conferences. Three typical scenarios have been identified, for each of which a reference (noiseless) speech signal has been selected, as well as a set of noises for these scenarios for conducting two types of tests in order to test the effectiveness of the noise reduction module in various conditions of its application. Experiments were carried out with a test of the effectiveness of noise reduction of noise segments during speech pauses using the methods of RMS signal volume and visual spectrogram analysis with calculation of the percentage of the number of samples of rhythms in the output signal. Experiments were carried out with a test of the effectiveness of noise reduction of a noisy speech signal using objective methods for evaluating the quality of the speech signal ViSQOL and NISQA. The test results are presented in the form of tables.

Sensorimotor environment but not task rule reconfigures population dynamics in rhesus monkey posterior parietal cortex

Primates excel at mapping sensory inputs flexibly onto motor outcomes. We asked if the neural dynamics to support context-sensitive sensorimotor mapping generalizes or differs between different behavioral contexts that demand such flexibility. We compared reaching under mirror-reversed vision, a case of adaptation to a modified sensorimotor environment (SE), with anti reaching, a case of applying an abstract task rule (TR). While neural dynamics in monkey posterior parietal cortex show shifted initial states and non-aligned low-dimensional neural subspaces in the SE task, remapping is achieved in overlapping subspaces in the TR task. A recurrent neural network model demonstrates how output constraints mimicking SE and TR tasks are sufficient to generate the two fundamentally different neural computational dynamics. We conclude that sensorimotor remapping to implement an abstract task rule happens within the existing repertoire of neural dynamics, while compensation of perturbed sensory feedback requires exploration of independent neural dynamics in parietal cortex.

Pediatric Readiness and Trauma Center Access for Children.

Children initially treated in a timely fashion at trauma centers with high levels of pediatric readiness have been shown to have improved survival, but children historically have had geographically disparate access to pediatric trauma center care. Considerable effort has been invested in improving pediatric readiness nationally, including the implementation of new standards to improve emergency department pediatric readiness at all trauma centers. To assess current access to US pediatric-ready trauma center care and to estimate potential improvement in access if all high-level trauma centers had optimal pediatric readiness. This descriptive cross-sectional study collated trauma centers from national organizational lists, state government websites, and online searches. A geospatial analysis was performed of access by pediatric patients (aged ≤18 years) to trauma centers by ground or air ambulance within 60 minutes, stratified by trauma center type and pediatric readiness status. Population density was estimated using 2020 US census data. Weighted pediatric readiness scores (wPRS) were obtained from the 2021 National Pediatric Readiness Project assessment. The data analysis was performed between April 1 and June 30, 2023. Access times to trauma centers. The main outcome of interest was access to a pediatric-ready trauma center, defined as a high-level pediatric trauma center (level I-II) or high-level adult trauma center (level I-III) with a wPRS of at least 93 (out of 100). Access times were calculated using previously validated methods and service network model analysis for each trauma center to census block group centroid. The analysis included 148 pediatric and 1075 high-level adult trauma centers. A total of 273 adult centers (25%) were pediatric ready. Pediatric trauma center access within 60 minutes by ground or air ambulance was available for 65% of all 74 090 665 children; 73% of children had access to a pediatric-ready trauma center within 60 minutes, and 92% had access to any high-level trauma center within 60 minutes. These findings suggest that access to pediatric trauma center care is limited, even with air ambulance transport. Ensuring pediatric readiness at all high-level adult trauma centers may substantially improve access to early high-quality initial resuscitative trauma care for children.

Predicting the hub interactome of COVID-19 and oral squamous cell carcinoma: uncovering ALDH-mediated Wnt/β-catenin pathway activation via salivary inflammatory proteins

Understanding shared pathways and mechanisms involved in the pathogenesis of diseases like oral squamous cell carcinoma (OSCC) and COVID-19 could lead to the development of novel therapeutic strategies and diagnostic biomarkers. This study aims to predict the interactome of OSCC and COVID-19 based on salivary inflammatory proteins. Datasets for OSCC and COVID-19 were obtained from https://www.salivaryproteome.org/differential-expression and selected for differential gene expression analysis. Differential gene expression analysis was performed using log transformation and a fold change of two. Hub proteins were identified using Cytoscape and Cytohubba, and machine learning algorithms including naïve Bayes, neural networks, gradient boosting, and random forest were used to predict hub genes. Top hub genes identified included ALDH1A1, MT-CO2, SERPINC1, FGB, and TF. The random forest model achieved the highest accuracy (93%) and class accuracy (84%). The naive Bayes model had lower accuracy (63%) and class accuracy (66%), while the neural network model showed 55% accuracy and class accuracy, possibly due to data pre-processing issues. The gradient boosting model outperformed all models with an accuracy of 95% and class accuracy of 95%. Salivary proteomic interactome analysis revealed novel hub proteins as potential common biomarkers.

Solving Multi-Criteria Shortest Path by Optimization with Morphological Filters

The challenge of determining the shortest path within a multimodal transportation network involves identifying the most efficient travel route while considering various interconnected modes of transportation, such as roads, railways, and public transit. This problem becomes increasingly complex when numerous criteria and modes are involved, complicating the decision-making process. This study proposes a novel approach to computing the shortest path in multimodal networks, focusing on four modes of transportation: metro, trams, buses, and taxis. The optimization criteria include distance, travel time, and monetary cost. The proposed method utilizes a new metaheuristic called Optimization by Morphological Filters (OMF), inspired by image processing techniques. This approach was compared with the Genetic Algorithms (GA) and the Non-Dominated Sorting Genetic Algorithm II (NSGA-II). Experiments were carried out using graph models of multimodal transport networks that closely resemble real-world scenarios varying in size. Furthermore, the proposed method was evaluated using a real network from the city of Lyon, France. The results demonstrate that the OMF approach performs well in terms of convergence to optimal solutions and computation time.

A Deep Learning with Metaheuristic Optimization-Driven Breast Cancer Segmentation and Classification Model using Mammogram Imaging

Cancer is the second leading cause of death globally, with Breast Cancer (BC) accounting for 20% of the new diagnoses, making it a major cause of morbidity and mortality. Mammography is effective for BC detection, but lesion interpretation is challenging, prompting the development of Computer-Aided Diagnosis (CAD) systems to assist in lesion classification and detection. Machine Learning (ML) and Deep Learning (DL) models are widely used in disease diagnosis. Therefore, this study presents an Optimized Graph Convolutional Recurrent Neural Network based Segmentation for Breast Cancer Recognition and Classification (OGCRNN-SBCRC) technique. In the preparation phase, images and masks are annotated and then classified as benign or malignant. To achieve this, the Wiener Filter (WF)-based noise removal and log transform-based contrast enhancement are used for preprocessing. The OGCRNN-SBCRC technique utilizes the UNet++ method for segmentation and the RMSProp optimizer for parameter tuning. In addition, the OGCRNN-SBCRC technique employs the ConvNeXtTiny Convolution Neural Network (CNN) approach for feature extraction. For BC classification and detection, the Graph Convolutional Recurrent Neural Network (GCRNN) model is used. Finally, the Aquila Optimizer (AO) model is employed for the hyperparameter tuning of the GCRNN approach. The simulation analysis of the OGCRNN-SBCRC methodology, using the BC image dataset, demonstrated superior performance with an accuracy of 99.65%, surpassing existing models.

Prediction of ketosis using radial basis function neural network in dairy cattle farming.

The purpose of the paper was to apply an Artificial Neural Networks with Radial Basis Function to develop an application model for diagnosing a subclinical ketosis type I and II in dairy cattle. While building the neural network model, applied methodology was compatible to the procedures used in Data Mining processes.The data set was created based on the composition of milk samples of 1520 Polish Holstein-Friesian cows. The milk samples were collected during test-day milkings and made available by Polish Federation of Cattle Breeders and Milk Producers. The milk composition parameters were used as the input variables for RBF network models. The value of the output variable was determined based on the content of β-hydroxybutyric acid in blood of cows. In the next stage of the work, the qualities of the pre-selected models were compared and the best ones were chosen. The sensitivity and specificity as well as the size of the AUC (Area Under the Curve) under the ROC (Receiver Operating Characteristic) were taken as the main criteria for network models evaluation. The model characterized by sensitivity of 0.86, specificity of 0.71 and AUC of 0.89 was selected for ketosis type I. The optimal for ketosis type II showed the sensitivity and specificity 0.81 and 0.75, respectively, and the size of AUC above 0.85. Chosen models were recorded using the predictive modelling markup language (PMML) for data mining models to be shared and used between the different applications.

Resonant song recognition and the evolution of acoustic communication in crickets.

Cricket song recognition is thought to evolve through modifications of a shared neural network. However, the species Anurogryllus muticus has an unusual recognition pattern that challenges this view: females respond to both normal male song pulse periods and periods twice as long. Of the three minimal models tested, only a single-neuron model with an oscillating membrane could explain this unusual behavior. A minimal model of the cricket's song network reproduced the behavior after adding a mechanism that, while present in the full network, is not crucial for song recognition in other species. This shows how a shared neural network can produce diverse behaviors and highlights how different computations contribute to evolution. Our results also demonstrate how nonlinear computations can lead to rapid behavioral changes during evolution because small changes in network parameters can lead to large changes in behavior.

Hybrid intrusion detection model for hierarchical wireless sensor network using federated learning

Hybrid intrusion detection model for hierarchical wireless sensor network using federated learning