Advanced Model Research Articles

Device Position-Independent Human Activity Recognition with Wearable Sensors Using Deep Neural Networks

Human activity recognition (HAR) identifies people’s motions and actions in daily life. HAR research has grown with the popularity of internet-connected, wearable sensors that capture human movement data to detect activities. Recent deep learning advances have enabled more HAR research and applications using data from wearable devices. However, prior HAR research often focused on a few sensor locations on the body. Recognizing real-world activities poses challenges when device positioning is uncontrolled or initial user training data are unavailable. This research analyzes the feasibility of deep learning models for both position-dependent and position-independent HAR. We introduce an advanced residual deep learning model called Att-ResBiGRU, which excels at accurate position-dependent HAR and delivers excellent performance for position-independent HAR. We evaluate this model using three public HAR datasets: Opportunity, PAMAP2, and REALWORLD16. Comparisons are made to previously published deep learning architectures for addressing HAR challenges. The proposed Att-ResBiGRU model outperforms existing techniques in accuracy, cross-entropy loss, and F1-score across all three datasets. We assess the model using k-fold cross-validation. The Att-ResBiGRU achieves F1-scores of 86.69%, 96.23%, and 96.44% on the PAMAP2, REALWORLD16, and Opportunity datasets, surpassing state-of-the-art models across all datasets. Our experiments and analysis demonstrate the exceptional performance of the Att-ResBiGRU model for HAR applications.

TransVAE-PAM: A Combined Transformer and DAG-based Approach for Enhanced Fake News Detection in Indian Context

In this study, we introduce a novel method, “TransVAE-PAM”, for the classification of fake news articles, tailored specifically for the Indian context. The approach capitalizes on state-of-the-art contextual and sentence transformer-based embedding models to generate article embeddings. Furthermore, we also try to address the issue of compact model size. In this respect, we employ a Variational Autoencoder (VAE) and β -VAE to reduce the dimensions of the embeddings, thereby yielding compact latent representations. To capture the thematic essence or important topics in the news articles, we use the Pachinko Allocation Model (PAM) model, a Directed Acyclic Graph (DAG) based approach, to generate meaningful topics. These two facets of representation - the reduced-dimension embeddings from the VAE and the extracted topics from the PAM model - are fused together to create a feature set. This representation is subsequently channeled into five different methods for fake news classification. Furthermore, we use eight distinct transformer-based architectures to test the embedding generation. To validate the feasibility of the proposed approach, we have conducted extensive experimentation on a proprietary dataset. The dataset is sourced from “Times of India” and other online media. Considering the size of the dataset, large-scale experiments are conducted on an NVIDIA supercomputer. Through this comprehensive numerical investigation, we have achieved an accuracy of 96.2% and an F1 score of 96% using the DistilBERT transformer architecture. By complementing the method via topic modeling, we record a performance improvement with the accuracy and F1 score both at 97%. These results indicate a promising direction toward leveraging the combination of advanced topic models into existing classification schemes to enhance research on fake news detection.

Decoupled representation for multi-view learning

Learning multi-view data is a central topic for advanced deep model applications. Existing efforts mainly focus on exploring shared information to maximize the consensus among all the views. However, after reasonably discarding superfluous task-irrelevant noise, the view-specific information is equally essential to downstream tasks. In this paper, we propose to decouple the multi-view representation learning into the shared and specific information extractions with parallel branches, and seamlessly adopt feature fusion in end-to-end models. The common feature is obtained based on the view-agnostic contrastive learning and view-discriminative training to minimize the discrepancy within the views. Simultaneously, the specific feature is learned with orthogonality constraints to minimize the view-level correlation. Besides, the semantic information in the features is reserved with supervised training. After disentangling the representations, we fuse the mutually complementary common and specific features for downstream tasks. Particularly, we provide a theoretical explanation for our method from an information bottleneck perspective. Compared with state-of-the-art multi-view models on benchmark datasets, we empirically demonstrate the advantage of our method in several downstream tasks, such as ordinary classification and few-shot learning. In addition, extensive experiments validate the robustness and transferability of our approach, when applying the learned representation on the source dataset to several target datasets.

MobileNet-GRU fusion for optimizing diagnosis of yellow vein mosaic virus

Yellow vein mosaic virus (YVMV) is a destructive plant virus that commonly affects crops, particularly okra, in India. The virus is transmitted by whiteflies and poses significant challenges to agricultural productivity. Infection with YVMV leads to distinct yellow vein patterns on leaves, stunted growth, reduced yield, and ultimately economic losses for farmers. Timely and accurate detection of YVMV is crucial for effective disease management. In this article, we present a novel method that employs advanced deep-learning models to identify YVMV-infected okra plants. The study leverages a dataset of over 2000 okra plant leaves that implements transfer learning models, including MobileNet, EfficientNet, InceptionV3, VGG19, InceptionResNetV2, and ResNet50 and recurrent neural networks (RNN) variants, including Long short-term memory (LSTM), Bidirectional long short-term memory (BiLSTM) and Gated recurrent unit (GRU). Additionally, three hybrid models, combining MobileNet with LSTM, BiLSTM, and GRU, are incorporated to capitalize on the characteristics of both MobileNet and RNNs through superior feature extraction and detection of temporal dependencies. The results demonstrate that the MobileNet model combined with all three RNNs achieves exceptional accuracy, surpassing 99.27%. Notably, the MobileNet model integrated with GRU exhibits the most optimized performance with the least loss and greatest accuracy, facilitating improved disease management strategies and aiding in the yield of crops by reducing the impact of YVMV.

Deep neural networks optimization for resource-constrained environments: techniques and models

This paper aims to present a comprehensive review of advanced techniques and models with a specific focus on deep neural network (DNN) for resource-constrained environments (RCE). The paper contributes by highlighting the RCE devices, analyzing challenges, reviewing a broad range of optimization techniques and DNN models, and offering a comparative assessment. The findings provide potential optimization techniques and recommend a baseline model for future development. It encompasses a broad range of DNN optimization techniques, including network pruning, weight quantization, knowledge distillation, depthwise separable convolution, residual connections, factorization, dense connections, and compound scaling. Moreover, the review analyzes the established optimization models which utilizes the above optimization techniques. A comprehensive analysis is conducted for each technique and model, considering its specific attributes, usability, strengths, and limitations in the context of effective deployment in RCEs. The review also presents a comparative assessment of advanced DNN models’ deployment for image classification, employing key evaluation metrics such as accuracy and efficiency factors like memory and inference time. The article concludes with the finding that combining depthwise separable convolution, weight quantization, and pruning represents potential optimization techniques, while also recommending EfficientNetB1 as a baseline model for the future development of optimization models in RCE image classification.<p> </p>

Revitalizing Cantonese Proficiency: An Interactive ASR-Driven Approach

In recent years, an increasing number of children in the Guangdong region found themselves unfamiliar with Cantonese which is their native language. This trend poses a significant threat to the preservation of the rich cultural heritage and traditions associated with the Cantonese language, meanwhile Cantonese is very different from Mandarin so it is very difficult to learn. Recognizing this challenge, the "Assistant for Cantonese Learning" app has been developed to bridge this gap. The app aids in using an ASR model to provide exercises tailored to a student's proficiency, allowing the user's mastery of Cantonese to be assessed based on the comparison between the translated text and the original text provided by the system, as well as the correctness of the user's pronunciation and intonation. Using OpenAI's advanced Whisper ASR model, the app offers an interactive and dynamic learning experience, ensuring that anyone who wants to learn Cantonese can connect with the linguistic roots efficiently and effectively.

Optical, thermal and semiempirical study of Samarium bi-tartrate trihydrate

In this comprehensive study, we investigated the structure, optical, and thermal aspects of [Sm(C4H5O6)(C4H4O6)][3H2O], a samarium based metal-organic framework. Testing and exploring advanced semiempirical quantum models, we provide some insights into ground-state molecular properties within the solid metal-organic framework. The single-excitation configuration interaction approach within the spectroscopic method INDO/S-CIS was utilized in the computation of UV spectra, singlet, and triplet excited states for the uncoordinated organic component. We also exploited the established Judd-Ofelt theory to elucidate the spectroscopic properties, allowing us to predict various attributes, including absorption and luminescence traits. The inclusion of thermogravimetric analysis exposes a multifaceted thermal decomposition process. Moreover, this analysis was explored in the determination of activation energies associated with distinct stages of decomposition.

Exploration of Machine Learning Model for Diabetes Prediction

Diabetes is a metabolic disorder characterized by a malfunction in insulin release, resulting in a rise in blood sugar levels in the body. Diabetes diagnosis must be made on time and precisely in order to be effective and enhance patient outcomes. The diabetes management presents a formidable challenge in modern healthcare, demanding a combination of timely interventions, precise data analysis, and personalized medical services. Furthermore, there exists a growing demand for advanced predictive models that not only provide accurate forecasts but also offer transparency and interpretability. The study objectives are to develop an innovative machine learning model for data-driven diabetes prediction and medical services. To identify the machine learning model that best suits the proposed system, a literature review related to different methods used in diagnosing diabetes is conducted. The merits and demerits of each existing system were identified to devise a proposed model that seamlessly integrates data from various sources, including blood glucose levels and patient health information. Based on the review, the study suggests an innovative machine learning model that utilizes the Explainable Decision Tree Model, leveraging cloud computing to analyse diabetes patient data and make predictions. The integration of cloud computing allows for seamless data integration from various sources. This research project represents a significant step forward in personalized diabetes care, enabling patients to proactively manage their condition while providing healthcare professionals with a powerful tool for delivering tailored medical services.

New beam-based models for fire-induced buckling analysis of class 4 steel columns

Steel cross-sections with thin walls are vulnerable to fire-induced buckling instability, which reduces their load-bearing capacity. Eurocode 3 design provisions have been found inadequate, leading to alternative methods such as effective design strategies and advanced structural models built mostly with shell FE, which can be complex. For Class 4 steel beam-columns subjected to fire conditions, beam-type modelling to predict the Flexural-Torsional Buckling (FTB) strength has been proposed as an alternative approach, but it has not yielded satisfactory results for large compressive load eccentricities. This paper presents two new low computational cost modelling strategies based on Timoshenko's beam FE to address this issue: the Single beam-column Model (SbcM) and the Cruciform beam-column Model (CbcM). The first consists of a single line of beam FE, while the second uses a grid of beam FE for more flexibility. Both strategies effectively simulate the FTB behaviour in Class 4 steel beam-column during a fire, offering quicker computations compared to shell models. Still, the single-line model is favoured for its simplicity, making it more efficient in analysing complex fire engineering problems.

Advanced Techniques for Geospatial Referencing in Online Media Repositories

In the digital transformation era, video media libraries’ untapped potential is immense, restricted primarily by their non-machine-readable nature and basic search functionalities limited to standard metadata. This study presents a novel multimodal methodology that utilizes advances in artificial intelligence, including neural networks, computer vision, and natural language processing, to extract and geocode geospatial references from videos. Leveraging the geospatial information from videos enables semantic searches, enhances search relevance, and allows for targeted advertising, particularly on mobile platforms. The methodology involves a comprehensive process, including data acquisition from ARD Mediathek, image and text analysis using advanced machine learning models, and audio and subtitle processing with state-of-the-art linguistic models. Despite challenges like model interpretability and the complexity of geospatial data extraction, this study’s findings indicate significant potential for advancing the precision of spatial data analysis within video content, promising to enrich media libraries with more navigable, contextually rich content. This advancement has implications for user engagement, targeted services, and broader urban planning and cultural heritage applications.

Data Augmentation Techniques for Accurate Action Classification in Stroke Patients with Hemiparesis.

Stroke survivors with hemiparesis require extensive home-based rehabilitation. Deep learning-based classifiers can detect actions and provide feedback based on patient data; however, this is difficult owing to data sparsity and heterogeneity. In this study, we investigate data augmentation and model training strategies to address this problem. Three transformations are tested with varying data volumes to analyze the changes in the classification performance of individual data. Moreover, the impact of transfer learning relative to a pre-trained one-dimensional convolutional neural network (Conv1D) and training with an advanced InceptionTime model are estimated with data augmentation. In Conv1D, the joint training data of non-disabled (ND) participants and double rotationally augmented data of stroke patients is observed to outperform the baseline in terms of F1-score (60.9% vs. 47.3%). Transfer learning pre-trained with ND data exhibits 60.3% accuracy, whereas joint training with InceptionTime exhibits 67.2% accuracy under the same conditions. Our results indicate that rotational augmentation is more effective for individual data with initially lower performance and subset data with smaller numbers of participants than other techniques, suggesting that joint training on rotationally augmented ND and stroke data enhances classification performance, particularly in cases with sparse data and lower initial performance.

Dynamic Modeling Framework Based on Automatic Identification of Operating Conditions for Sintering Carbon Consumption Prediction

In iron and steel industry, sintering process requires huge carbon consumption. Achieving accurate and dynamic prediction of carbon consumption in this process is of evident significance to protect the environment and raise the economic efficiency of iron and steel industry. This paper develops an original dynamic modeling framework including automatic identification of operating conditions, modeling in different conditions, and dynamic prediction model of sintering carbon consumption. Firstly, an automatic kernel-based fuzzy C-means algorithm is presented for automatic identification of operating conditions. Dynamic relationships between the inputs and output of the model are analyzed by Copula Entropy, and then the relevant production data in each operating condition are determined by just-in-time learning method. Next, broad learning models are established under different operating conditions. Further, dynamic prediction model of sintering carbon consumption is designed, and the prediction error is adopted to quantify performance of the model and as one criteria to determine the update of production database. Finally, results of experiments using actual production data demonstrate the advantage and validity of the developed model compared with some advanced modeling methods. The developed model considers complex characteristics of the sintering process and presents a better dynamic performance and information mining performance.

Bifurcation analysis of thin-walled structures trimming process with state-dependent time delay

Our previous study [1] revealed that trimming of thin-walled structures has a strongly nonlinear nature, hence this investigation delves into nonlinear dynamic analysis. Specifically, a perturbation method was applied to analyze the linear stability, whilst a direct numerical integration approach was employed to construct bifurcation diagrams, where unsafe cutting caused by multiple stabilities had been found. Basins of attraction were computed to evaluate the cutting safety. Finally, the stability charts for both linearized and fully nonlinear models were constructed showing a good degree of similarity but the linear one has a larger area of stable trimming. The data from our focused experimental studies was then superimposed on both diagrams and it is clear that more refinements in models’ calibration are needed. This work has laid firm foundations for further advanced modelling investigations and experimental studies to gain more insights into nonlinear dynamics of the machining process.

Recent Advances in Thermal Management Strategies for Lithium-Ion Batteries: A Comprehensive Review

Effective thermal management is essential for ensuring the safety, performance, and longevity of lithium-ion batteries across diverse applications, from electric vehicles to energy storage systems. This paper presents a thorough review of thermal management strategies, emphasizing recent advancements and future prospects. The analysis begins with an evaluation of industry-standard practices and their limitations, followed by a detailed examination of single-phase and multi-phase cooling approaches. Successful implementations and challenges are discussed through relevant examples. The exploration extends to innovative materials and structures that augment thermal efficiency, along with advanced sensors and thermal control systems for real-time monitoring. The paper addresses strategies for mitigating the risks of overheating and propagation. Furthermore, it highlights the significance of advanced models and numerical simulations in comprehending long-term thermal degradation. The integration of machine learning algorithms is explored to enhance precision in detecting and predicting thermal issues. The review concludes with an analysis of challenges and solutions in thermal management under extreme conditions, including ultra-fast charging and low temperatures. In summary, this comprehensive review offers insights into current and future strategies for lithium-ion battery thermal management, with a dedicated focus on improving the safety, performance, and durability of these vital energy sources.

Interactive Human 3D Model with Conventional AI

Abstract: The convergence of interactive human 3D models with conventional artificial intelligence (AI) represents a groundbreaking fusion of computer graphics and intelligent systems. This interdisciplinary field combines advanced 3D modeling techniques with classical AI methodologies to create interactive virtual representations of humans. Rigorous studies in 3D modeling and animation form the basis for lifelike avatars, while conventional AI empowers these models with the ability to understand user inputs, make contextually relevant decisions, and engage in dynamic and responsive interactions. At the heart of this innovation is the emphasis on human-computer interaction (HCI), leveraging principles such as gesture recognition, voice commands, and user-friendly interfaces to enhance user experiences. The integration extends into virtual and augmented reality (VR/AR), providing users with immersive and engaging environments where interactive 3D human models come to life. Applications of this technology span a wide array of fields, including virtual companionship, education, healthcare simulations, entertainment, and training scenarios. The impact is transformative, offering new possibilities for human-machine interactions and redefining traditional approaches to AI applications. However, the burgeoning field is not without challenges. Privacy concerns, ethical considerations related to realistic human representation, and the societal impact of interactive 3D models necessitate careful examination. This abstract provides a glimpse into the technical intricacies of creating lifelike models, the intelligence driving their interactive capabilities, and the broader implications for diverse applications and industries. As research and development progress, the collaboration of computer graphics and AI not only opens new avenues but also prompts a thoughtful reevaluation of the evolving relationship between humans and intelligent, interactive 3D representations.

Comprehensive Survey on Disease Detection in Retinal Imaging through Deep Learning Techniques

Abstract: Millions of people around the world suffer from visual impairment and blindness that could have been prevented or diagnosed earlier. To tackle this issue, we’ve introduced a groundbreaking approach to detect multiple eye diseases using retinal imaging. Our study involves a unique combination of various advanced deep learning models through ensemble learning.We present a detailed examination of deep learning models in the context of multi-disease classification using the Retinal Fundus MultiDisease Image Dataset (RFMiD). The objective is to assess the performance of these models and identify the most effective one for accurate and reliable disease diagnosis.We begin by preprocessing the RFMiD dataset to enhance image quality and extract relevant features, ensuring a robust input for our models. Subsequently, each model undergoes a comprehensive training and finetuning process to optimize its parameters for disease classification. The evaluation metrics include accuracy, precision, recall, and F1 score, providing a comprehensive understanding of model performance.We discuss the strengths and weaknesses of each model, shedding light on factors influencing their performance. The insights gained from this comparative analysis can guide researchers and practitioners in selecting an appropriate model for retinal disease diagnosis based on specific requirements and constraints.In addition, we explore potential avenues for future research, including ensemble methods and hybrid architectures, to further improve classification accuracy. The paper includes discussion on the practical implications of our findings and their significance in advancing the field of medical image analysis, particularly in the context of retinal disease diagnosis.

Pixelwise Gradient Model with GAN for Virtual Contrast Enhancement in MRI Imaging.

Background: The development of advanced computational models for medical imaging is crucial for improving diagnostic accuracy in healthcare. This paper introduces a novel approach for virtual contrast enhancement (VCE) in magnetic resonance imaging (MRI), particularly focusing on nasopharyngeal cancer (NPC). Methods: The proposed model, Pixelwise Gradient Model with GAN for Virtual Contrast Enhancement (PGMGVCE), makes use of pixelwise gradient methods with Generative Adversarial Networks (GANs) to enhance T1-weighted (T1-w) and T2-weighted (T2-w) MRI images. This approach combines the benefits of both modalities to simulate the effects of gadolinium-based contrast agents, thereby reducing associated risks. Various modifications of PGMGVCE, including changing hyperparameters, using normalization methods (z-score, Sigmoid and Tanh) and training the model with T1-w or T2-w images only, were tested to optimize the model's performance. Results: PGMGVCE demonstrated a similar accuracy to the existing model in terms of mean absolute error (MAE) (8.56 ± 0.45 for Li's model; 8.72 ± 0.48 for PGMGVCE), mean square error (MSE) (12.43 ± 0.67 for Li's model; 12.81 ± 0.73 for PGMGVCE) and structural similarity index (SSIM) (0.71 ± 0.08 for Li's model; 0.73 ± 0.12 for PGMGVCE). However, it showed improvements in texture representation, as indicated by total mean square variation per mean intensity (TMSVPMI) (0.124 ± 0.022 for ground truth; 0.079 ± 0.024 for Li's model; 0.120 ± 0.027 for PGMGVCE), total absolute variation per mean intensity (TAVPMI) (0.159 ± 0.031 for ground truth; 0.100 ± 0.032 for Li's model; 0.153 ± 0.029 for PGMGVCE), Tenengrad function per mean intensity (TFPMI) (1.222 ± 0.241 for ground truth; 0.981 ± 0.213 for Li's model; 1.194 ± 0.223 for PGMGVCE) and variance function per mean intensity (VFPMI) (0.0811 ± 0.005 for ground truth; 0.0667 ± 0.006 for Li's model; 0.0761 ± 0.006 for PGMGVCE). Conclusions: PGMGVCE presents an innovative and safe approach to VCE in MRI, demonstrating the power of deep learning in enhancing medical imaging. This model paves the way for more accurate and risk-free diagnostic tools in medical imaging.

Failure Pressure Evaluation of Corroded Pipeline Using Semi-Empirical and Finite Element Analysis

Abstract Computational simulation using the finite element method (FEM) has proven to be one of the most efficient methods for the correct evaluation of the structural integrity of pipelines with corrosion-induced defects. In this paper, the results of the failure pressure of corroded pipelines obtained with a computer tool called PIPEFLAW based on nonlinear finite element analyses are presented and compared with the failure pressures of corroded pipes obtained with twelve (12) advanced semi-empirical models from the literature. A computational code was created in the Python language to automatically obtain the failure pressure of randomly generated cases of corroded pipes, which allows the analysis of multiple piping cases with isolated external corrosion defects in an idealized configuration. For this purpose, the LHS (Latin Hypercube Sampling) approach was adopted for the design of experiments to obtain a better distribution of cases in the sample space. Comparisons between the results using the FEM and semi-empirical models show, as expected, that the semi-empirical solutions are more conservative in most cases analyzed in this work. A statistical model is proposed that adheres to the relative error distribution for some of the models used. Moreover, the influence of some parameters (pipeline thickness, defect depth, defect length, defect width, yield stress and ultimate stress) on the failure pressure was evaluated. It was found that in most of the studied models the error increases as the defect depth increases.

Enhanced Crowd Dynamics Simulation with Deep Learning and Improved Social Force Model

The traditional social force model (SFM) in crowd simulation experiences difficulty coping with the complexity of the crowd, limited by singular physical formulas and parameters. Recent attempts to combine deep learning with these models focus more on simulating specific states of crowds. This paper introduces an advanced deep social force model, influenced by crowd states. It utilizes deep neural networks to accurately fit crowd trajectory features, enhancing behavior simulation capabilities. Geometrical constraints within the model provide control over varied crowd behaviors, adjustable to simulate different crowd types. Before training, we use the SFM to refine behaviors in real trajectories with excessively small distances, aiming to enhance the general applicability of the model. Comparative experiments affirm the effectiveness of the model, showing comparable performance to both classic physical models and modern learning-based hybrid models in pedestrian simulations, with reduced collisions. In addition, the model has a certain ability to simulate crowds with high density and diverse behaviors.

Border Defence Mechanism Using Deep Learning Technique

In response to escalating concerns regarding potential border airstrikes, this paper introduces an advanced deep-learning model that utilizes Convolutional Neural Networks (CNNs) for accurate military aircraft classification. Functioning as a pivotal tool for proactive threat assessment, the system works by distinguishing and categorizing six aircraft types with exceptional accuracy, facilitating real-time analysis, and aiding with the development of responsive counter-attack strategies. The multi-stage process begins with the comprehensive collection of diverse datasets, followed by the extraction of spatial features from images using CNNs. Optimization techniques fine-tune model parameters to ensure optimal performance. This approach significantly strengthens national security by leveraging advanced technology for proactive threat mitigation. Emphasizing this paper’s vital role in enhancing border defence capabilities, it highlights the model’s capacity to navigate evolving security challenges through improved awareness of the environment. The model's ability to adapt to real-world conditions, including variations in lighting, weather, and terrain, demonstrates its practical applicability. For seamless implementation, the deep learning model is deployed using Django and SQLite as a web page providing an efficient and user-friendly interface This sophisticated deep learning model is fundamental in fortifying national security measures against potential threats along border regions.