Advanced Architectures Research Articles

Advancing Prostate Cancer Diagnosis: A Deep Learning Approach for Enhanced Detection in MRI Images.

Prostate cancer remains a leading cause of mortality among men globally, necessitating advancements in diagnostic methodologies to improve detection and treatment outcomes. Magnetic Resonance Imaging has emerged as a crucial technique for the detection of prostate cancer, with current research focusing on the integration of deep learning frameworks to refine this diagnostic process. This study employs a comprehensive approach using multiple deep learning models, including a three-dimensional (3D) Convolutional Neural Network, a Residual Network, and an Inception Network to enhance the accuracy and robustness of prostate cancer detection. By leveraging the complementary strengths of these models through an ensemble method and soft voting technique, the study aims to achieve superior diagnostic performance. The proposed methodology demonstrates state-of-the-art results, with the ensemble model achieving an overall accuracy of 91.3%, a sensitivity of 90.2%, a specificity of 92.1%, a precision of 89.8%, and an F1 score of 90.0% when applied to MRI images from the SPIE-AAPM-NCI PROSTATEx dataset. Evaluation of the models involved meticulous pre-processing, data augmentation, and the use of advanced deep-learning architectures to analyze the whole MRI slices and volumes. The findings highlight the potential of using an ensemble approach to significantly improve prostate cancer diagnostics, offering a robust and precise tool for clinical applications.

Urban air quality index forecasting using multivariate convolutional neural network based customized stacked long short-term memory model

In the context of increasing urban pollution and its adverse health effects, this study focuses on enhancing the air quality prediction model by forecasting future concentrations of various Air Quality Index (AQI) levels to support the development of green smart cities. The primary objective of this study is to develop a robust multivariate time series forecasting model using a combination of Convolutional Neural Networks (CNN) and Customized Stacked-based Long Short-Term Memory (CSLSTM) networks, namely C2SLSTM. The proposed methodology involved preprocessing AQI and meteorological data, scaling the features, and employing a hybrid CNN-LSTM architecture to capture spatial and temporal dependencies in the data. Dynamic AQI pattern changes are computed using online learning. The proposed model was trained on a big dataset containing AQIs such as particulate matter (PM2.5), carbon monoxide (CO), and nitrogen dioxide (NO2), along with meteorological factors like weather count, temperature, wind, humidity, and pressure. Experimental results demonstrate that the model achieved a mean absolute error (MAE) of 0.25, mean square error (MSE) of 0.083, root mean square error (RMSE) of 0.289 and a R2-Score of 0.84 on the test set. Also, compared to Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), Stacked Long Short-Term Memory (SLSTM), and Multilayer Perceptron (MLP) models, the C2SLSTM model showcases a performance enhancement of 35 %, 50 %, 30 %, and 32 %. Notably, the CNN component effectively extracted spatial features, while the LSTM layers captured temporal patterns, leading to precise predictions. These results indicate that the proposed approach effectively captures temporal dependencies and improves forecasting performance, with R2-Score values tending towards 1 for 12-hour prediction intervals. This research demonstrates the potential of advanced SLSTM architectures in accurately predicting AQI metrics, which can aid in better environmental monitoring and management. Also, the spatiotemporal correlation analysis is done through the designed ir metric.

IoT-Based Assessment of a Driver's Stress Level.

Driver Monitoring Systems (DMSs) play a key role in preventing hazardous events (e.g., road accidents) by providing prompt assistance when anomalies are detected while driving. Different factors, such as traffic and road conditions, might alter the psycho-physiological status of a driver by increasing stress and workload levels. This motivates the development of advanced monitoring architectures taking into account psycho-physiological aspects. In this work, we propose a novel in-vehicle Internet of Things (IoT)-oriented monitoring system to assess the stress status of the driver. In detail, the system leverages heterogeneous components and techniques to collect driver (and, possibly, vehicle) data, aiming at estimating the driver's arousal level, i.e., their psycho-physiological response to driving tasks. In particular, a wearable sensorized bodice and a thermal camera are employed to extract physiological parameters of interest (namely, the heart rate and skin temperature of the subject), which are processed and analyzed with innovative algorithms. Finally, experimental results are obtained both in simulated and real driving scenarios, demonstrating the adaptability and efficacy of the proposed system.

A Survey an Optimized Dense CNN Model for Recognizing Deepfake Images

Deepfake detection relies on a deep learning model. Deepfake content is generated using artificial intelligence and machine learning techniques to swap one person’s face with another’s in images. These altered images are certain to have a significant impact on society. Deepfakes utilize advanced technologies such as machine learning (ML) and deep learning (DL) to develop automated techniques for generating deceptive content. The model is trained using the DFDC (Deepfake Detection Challenge) dataset, which includes 1, 00,000 videos comprising both real and fake content. This survey examines recent progress in deep learning techniques for detecting deepfake images, highlighting their growing significance in today’s digital landscape. The paper outlines the issues arising from deepfakes, explores the techniques for generating deceptive content, and analyzes various deep learning models employed in deepfake detection, such as convolutional neural networks (CNN’s), MTCNN (Multi-task CNN), and Facial Landmark. It emphasizes the ongoing battle between the progression of deepfakes and the evolution of detection methods, underscoring the need for advanced and flexible neural network architectures to effectively curb the dissemination of deceptive information.

Multi-modal transformer architecture for medical image analysis and automated report generation

Medical practitioners examine medical images, such as X-rays, write reports based on the findings, and provide conclusive statements. Manual interpretation of the results and report generation by examiners are time-consuming processes that lead to potential delays in diagnosis. We propose an automated report generation model for medical images leveraging an encoder–decoder architecture. Our model utilizes transformer architectures, including Vision Transformer (ViT) and its variants like Data Efficient Image Transformer (DEiT) and BERT pre-training image transformer (BEiT), as an encoder. These transformers are adapted for processing to extract and gain visual information from medical images. Reports are transformed into text embeddings, and the Generative Pre-trained Transformer (GPT2) model is used as a decoder to generate medical reports. Our model utilizes a cross-attention mechanism between the vision transformer and GPT2, which enables it to create detailed and coherent medical reports based on the visual information extracted by the encoder. In our model, we have extended the report generation with general knowledge, which is independent of the inputs and provides a comprehensive report in a broad sense. We conduct our experiments on the Indiana University X-ray dataset to demonstrate the effectiveness of our models. Generated medical reports from the model are evaluated using word overlap metrics such as Bleu scores, Rouge-L, retrieval augmentation answer correctness, and similarity metrics such as skip thought cs, greedy matching, vector extrema, and RAG answer similarity. Results show that our model is performing better than the recurrent models in terms of report generation, answer similarity, and word overlap metrics. By automating the report generation process and incorporating advanced transformer architectures and general knowledge, our approach has the potential to significantly improve the efficiency and accuracy of medical image analysis and report generation.

Integrating Artificial Intelligence into the Supply Chain in Order to Enhance Sustainable Production—A Systematic Literature Review

Nowadays, integrating Artificial Intelligence (AI) into supply chains (SCs) is a great challenge in research and for manufacturing managers. The main goal of this study is to determine the role of AI in the context of the new SCs, according to the concept of Industry 5.0. in order to improve the level of sustainable production. The research was based on a systematic analysis of the scientific literature and application of the PRISMA methodology. Due to the relatively new vision of introducing AI into SC, it was decided to analyse the years 2021–2024. A total of 1181 research articles were identified in Science Direct, Springer and the Willey Online Library that combined AI-based methods and tools that support SCs in order to identify the impacts and challenges of integrating AI in SCs in the context of sustainable production (SP). In this study, 48 items were then analysed in detail. The results achieved highlighted the main AI-based tools applied in SCs and, secondly, revealed the main benefits of this integration for manufacturing in the following areas of manufacturing: predictive maintenance, production planning and customer relationships. The findings of our study revealed the main challenges and directions: (1) integrating digitalisation and green SP in order to build resilience to the SP, (2) create a sustainable work environment, (3) and develop a sustainable and advanced architecture for relationships with customers.

Automated detection of exterior cladding material in urban area from street view images using deep learning

Street View Images (SVIs) can be utilized in urban analysis to assess the exterior cladding materials of buildings. However, this process is labor-intensive and time-consuming, particularly when applied to extensive geographic areas. Currently, there is no automated study to streamline this manual process. This research introduces a deep learning approach suitable for image classification and demonstrates its application in two case study areas: London and Scotland. Six material types were identified: ‘Brick’, ‘Concrete’, ‘Glass’, ‘Stone’, ‘Mixed’, and ‘Others’. To ensure reliable accuracy, several advanced architectures were employed: InceptionV3, EfficientNetV2, ResNet-101, ResNet-152, and MobileNetV3. Transfer learning was applied to each architecture, and six distinct image augmentation techniques were utilized to artificially expand the training dataset. Consequently, ten models were developed for each area by combining five different architectures with two datasets: the original and the augmented. The best-performing model was selected for evaluation on an unseen dataset. For London and Scotland, the model using MobileNetV3 with augmented data emerged as the most effective, achieving average accuracy rates of 75.65 % and 73.45 %, respectively. Based on these findings, the paper explores the potential applications of the proposed approach.

Mitigating digital market risk with conventional, green, and Islamic bonds: Fresh insights from new hybrid deep learning models

We examine the impact of conventional, green, and Islamic bonds on the long-term memory of cryptocurrency market risk. Utilizing a time-varying parameter vector autoregressive deep learning model, we integrate time-varying parameter vector autoregressive methods with advanced deep learning sequence modeling architectures, including temporal convolutional network, gated recurrent unit, and long short-term memory for December 18, 2017, to April 19, 2024. Results indicate that incorporating all fixed-income securities reduces digital market risk. However, conventional and green bonds have a particularly strong impact on improving the long-term memory of digital market risk, while this is not the case for Sukuk.

A revocable multi-authority attribute-based encryption scheme for fog-enabled IoT

As more applications move data storage to the cloud, protecting sensitive data becomes increasingly important, especially for the Internet of Things (IoT) environments. Ciphertext-policy attribute-based encryption (CP-ABE) is a practical approach for confidentiality and secure access control for data outsourced to the cloud. However, the underlying CP-ABE operations based on bilinear pairings are too demanding for resource-constrained IoT devices. Furthermore, applications such as Industrial IoT (IIoT) have requirements (efficiency, operative, and security) that existing CP-ABE proposals cannot fulfill, and advanced IoT architectures (e.g., fog computing) have not been well-exploited. This paper proposes a novel CP-ABE scheme suitable for IoT scenarios, using an IIoT generic model as a reference. It targets multiple attribute authorities, outsourced encryption and decryption to fog nodes, user revocation, and asymmetric pairings constructions to achieve recommended security levels. As the main distinctive, revocation is defined using a broadcast encryption-based approach, allowing data owners to enforce user revocation over their outsourced data. According to the performance analysis, the proposed scheme achieves high efficiency for IoT nodes. It is also competitive in terms of storage, bandwidth, and computation efficiency compared to previous proposals. Moreover, the security of the suggested construction is demonstrated against chosen-plaintext attacks.

Leveraging Deep Learning for Time-Series Extrinsic Regression in Predicting the Photometric Metallicity of Fundamental-Mode RR Lyrae Stars.

Astronomy is entering an unprecedented era of big-data science, driven by missions like the ESA's Gaia telescope, which aims to map the Milky Way in three dimensions. Gaia's vast dataset presents a monumental challenge for traditional analysis methods. The sheer scale of this data exceeds the capabilities of manual exploration, necessitating the utilization of advanced computational techniques. In response to this challenge, we developed a novel approach leveraging deep learning to estimate the metallicity of fundamental mode (ab-type) RR Lyrae stars from their light curves in the Gaia optical G-band. Our study explores applying deep-learning techniques, particularly advanced neural-network architectures, in predicting photometric metallicity from time-series data. Our deep-learning models demonstrated notable predictive performance, with a low mean absolute error (MAE) of 0.0565, the root mean square error (RMSE) of 0.0765, and a high R2 regression performance of 0.9401, measured by cross-validation. The weighted mean absolute error (wMAE) is 0.0563, while the weighted root mean square error (wRMSE) is 0.0763. These results showcase the effectiveness of our approach in accurately estimating metallicity values. Our work underscores the importance of deep learning in astronomical research, particularly with large datasets from missions like Gaia. By harnessing the power of deep-learning methods, we can provide precision in analyzing vast datasets, contributing to more precise and comprehensive insights into complex astronomical phenomena.

Moore's Law Scaling and Chemical Mechanical Planarization (CMP)

Since the invention of the first integrated circuit in 1958, chemistry and materials have played a critical role in semiconductor process technologies, which has led to an exponential growth of transistors packed into circuits. Decades ago, Gordon Moore posited there would be a doubling of components every two years (revised from his original doubling every one-year prediction). This observation became known as Moore’s law.As Denard scaling took on more complexity, the number of new materials used to build semiconductors increased at a torrid pace. The number of elements in the periodic table, used to fabricate devices, increased roughly five-fold from the 1980’s to the 2000’s.In the earlier years, lithography, etch, thin films, implant, and diffusion were foundational technologies used to fabricate a silicon wafer into hundreds of microchips. In the late 1980’s a new technology Chemical Mechanical Planarization (CMP) was invented. It was developed out of necessity to reduce topography, critical for lithography’s depth of focus requirements. Without this key invention, patterning and building layer upon layer from front end transistors to back-end interconnects would not have been possible.Intel implemented CMP into manufacturing in the early 1990’s on the 0.8 micron technology node which manufactured the 486 microprocessor. CMP was required to planarize dielectric oxide used to insulate the aluminum metal interconnects. Since then, the number of CMP applications over the last three decades has exploded, enabling the scaling of Moore’s Law.CMP is not only used to improve lithography depth of focus. CMP also enables a method of metal interconnect and contact formation called damascene. Planarization is used to create dielectric isolation trenches, discrete plugs for insulation, aid advanced patterning architectures, build through silicon vias (TSVs), and enable complex wafer and chip bonding for die-to-die interconnects for disaggregated products.Transistor scaling alone is no longer sufficient to meet device performance, power, and cost requirements. Complex material and architectural changes are required. And novel CMP applications continue to be added at each new technology node. This author will make an argument that CMP is critical to push scaling and enable future process nodes. Figure 1

Bridging Technology and Environment: Sustainable Approaches in Semiconductor Chemical Mechanical Planarization (CMP)

As semiconductor technologies continue to advance in areas such as autonomous driving, artificial intelligence, 5G communications, the Internet of Things, and large-scale data processing, the demand for a reliable semiconductor industry is on the rise.(1) This growing demand necessitates the use of advanced device architectures, which heavily rely on two crucial processes: Chemical Mechanical Planarization (CMP) and post-CMP cleaning.(2) This talk will focus on the development of CMP slurries and post-CMP cleaning solutions, emphasizing their vital roles in these processes. We will delve into the comprehensive development of CMP slurries and post-CMP cleaning solutions, highlighting how synthesis methods can significantly affect the surface chemistry of ceria abrasives and impact the removal rates of SiO2 surfaces during STI CMP.(3, 4) Our research also demonstrates how our cleaning solutions can effectively remove even the most 10 nm-sized ceria particles from SiO2 surfaces,(5) thereby enhancing process efficiency. Furthermore, we will discuss the use of aliphatic amino acids as environmentally friendly corrosion inhibitors in CMP slurries, offering a sustainable alternative to the traditionally used benzotriazole (BTA) and addressing key environmental challenges.(6) In the context of the rapidly expanding semiconductor industry, it becomes critical to address global sustainability concerns and environmental health and safety (EHS) objectives. Our research proposes a comprehensive methodology for evaluating the sustainability of CMP consumables in semiconductor manufacturing, with a primary focus on CMP slurries due to their significant market share and shorter usage lifetimes. This research serves as a foundation for future CMP sustainability analyses, encouraging self-correction strategies within the CMP community to reduce environmental impact. We will also introduce a robust framework for the Life Cycle Assessment (LCA) of CMP consumables, marking a crucial step toward achieving these sustainability goals. In summary, this talk will explore the intricate details of CMP and post-CMP processes, their pivotal roles in the advancement of semiconductor manufacturing, and the imperative task of aligning these processes with environmental and sustainability objectives.This talk is dedicated to the special session in honor of Prof. Babu. S. Babu, Advances in Chemical Mechanical Planarization (CMP), Woodhead Publishing (2021).J. Seo, Journal of Materials Research, 36, 235 (2021).J. Seo, J. W. Lee, J. Moon, W. Sigmund and U. Paik, ACS applied materials & interfaces, 6, 7388 (2014).J. Seo, J. H. Kim, M. Lee, J. Moon, D. K. Yi and U. Paik, Journal of colloid and interface science, 483, 177 (2016).J. Seo, A. Gowda and S. Babu, ECS Journal of Solid State Science and Technology, 7, P243 (2018).H. T. T. Tran, J. Gichovi, E. J. Podlaha and J. Seo, in Electrochemical Society Meeting s 243, p. 1809 (2023).

Deciphering the Impact of Current, Composition, and Potential on the Lithiation Behavior of Graphite in Silicon-Graphite Anodes

Graphite is the most used anode material for current-generation lithium-ion batteries due to its beneficial cycle life, availability and reasonable rate capability. Still, the modest capacity of 372 mAhg-1 represents a bottleneck in the pursuit of high-energy density anodes. Enabling much higher theoretical capacities of 3600 mAhg-1 [1], silicon represents a promising candidate as anode material for next generation batteries. However, the alloying-based lithiation mechanism of silicon is accompanied by a large volumetric change of up to 300% upon cycling, leading to fast degradation due to electrode pulverization, continuous SEI formation and extensive electrolyte consumption. Even though shallow cycling of silicon anodes has been demonstrated in laboratory scale cells [2], the mixing of silicon with graphite is considered the most viable approach to lift energy density while maintaining appropriate cycle life.Due to the fundamentally different kinetics and potentials of Li insertion and extraction of both materials, the development of silicon-graphite composite electrodes is not straightforward. One of the key questions for optimization is how the incorporation of silicon impacts the (de)lithiation behavior of graphite as a function of the silicon:graphite mass ratio, operating current and electrode potential. While lithiation of graphite in composite electrodes of 15 wt% silicon has been shown to occur in a similar manner as pure graphite, higher fractions of silicon are expected to exert more stress on the graphite and represent kinetic barriers for Li diffusion. This study presents results from in-situ X-ray diffraction experiments (XRD) of silicon-graphite/Li half cells conducted at the European Synchrotron Radiation Facility (ESRF). Serving as a reference, the current-dependent (de)lithiation behavior of a pure graphite/Li half cell has been thoroughly investigated and systematically compared to the (de)lithiation behavior of two selected silicon-graphite composite anodes (silicon:graphite ratio of 30:70 and 70:30, respectively) exposed to the same formation cycle and C-rate protocol. An optimized and novel measurement setup was used, based on a perforated current collector of the anode, providing a complete picture of the graphite in-plane and interplanar structural changes as well as the evolution of semicrystalline silicon peaks which would usually be obscured by the current collector signal.By correlating the electrochemical features (voltage curve and differential capacity plot) with the in-situ diffraction data, it was possible to identify and assign the occurrence and absence of dilute-stage and ordered graphite intercalation compounds (GICs) for the respective electrodes, yielding an in-depth insight into the graphite state of charge upon (de)lithiation and how it is influenced by the silicon content and applied C-rate.As expected, the silicon is (de)lithiated within the entire potential range, whereas graphite is most electrochemically active at potentials lower than 260 mV. In both composite electrodes, graphite attains a lower lithiation degree compared to pure graphite. This is because the high theoretical capacity of silicon results in high specific currents, ultimately challenging the rate capability of graphite. Moreover, the high delithiation overpotential encountered in the composite electrodes results in a highly asymmetrical lithiation/delithiation behavior of graphite. Also, graphite lithiation is less uniform in the composite anodes, indicated by the coexistence of dilute GICs upon (de)lithiation, compared to pure graphite where dilute phases are fully consumed (formed) as (de)lithiation progresses. Surprisingly, the in-situ XRD studies did not reveal signs of structural degradation and strain of graphite as a result of mechanical interaction with silicon. Instead, the volume change of graphite is well correlated with the amorphization of crystalline, unreacted silicon and the volume evolution of silicon crystallites upon charge-discharge.To mitigate the observed effects, the work suggests nanostructuring and advanced electrode architectures towards higher utilization and more homogeneous lithiation of graphite. Overall, the study provides a demonstration of a suitable operando cell for studies of anodes for Li-based systems, and aids the rational design of silicon-graphite composite anodes.

Software Update Methodologies for Feature-Based Product Lines: A Combined Design Approach

The automotive industry is experiencing a significant shift, transitioning from traditional hardware-centric systems to more advanced software-defined architectures. This change is enabling enhanced autonomy, connectivity, safety, and improved in-vehicle experiences. Service-oriented architecture is crucial for achieving software-defined vehicles and creating new business opportunities for original equipment manufacturers. A software update approach that is rich in variability and based on a Merkle tree approach is proposed for new vehicle architecture requirements. Given the complexity of software updates in vehicles, particularly when dealing with multiple distributed electronic control units, this software-centric approach can be optimized to handle various architectures and configurations, ensuring consistency across all platforms. In this paper, our software update approach is expanded to cover the solution space of the feature-based product line engineering, and we show how to combine our approach with product line engineering in creative and unique ways to form a software-defined vehicle modular architecture. Then, we offer insights into the design of the Merkle trees utilized in our approach, emphasizing the relationship among the software modules, with a focus on their impact on software update performance. This approach streamlines the software update process and ensures that the safety as well as the security of the vehicle are continuously maintained.

Exploratory study on low-GWP working fluid mixtures for industrial high temperature heat pump with 200 °C supply temperature

The development of an industrial high temperature heat pump (HTHP) using a subcritical vapor-compression cycle (VCC) for process heat delivery at 200 °C faces significant challenges due to the scarcity of suitable single-component working fluids. These challenges are further compounded by safety, environmental concerns, material and lubrication oil compatibility and cost considerations. In response, this study investigates the potential of various fluid mixtures to meet these stringent criteria. Using the REFPROP and Cantera software libraries, a Python code was developed to simulate a simplified HTHP VCC with an advanced cycle architecture and assess the flammability risks of the proposed mixtures. Ten single-component fluids of two different families of compounds, hydrocarbons (HCs) and hydro(chloro)fluoroolefins (H(C)FOs), were pre-selected as promising candidates as mixture components. From these, over 460,000 binary, ternary, and quaternary mixtures were generated and evaluated for cycle performance. Although no mixture perfectly met all the predefined criteria, a binary blend of cyclopentane and R1336mzz(Z) in a mole fraction of 0.68 and 0.32, respectively, emerged as the most promising candidate. This mixture demonstrated the best balance of performance and safety under the imposed constraints and is selected for further experimental investigation and validation in a specially designed test rig.

FusionEXNet: an interpretable fused deep learning model for skin cancer detection

Skin cancer poses a significant burden on mankind and healthcare systems globally, necessitating the development of effective diagnostic and treatment strategies. This paper introduces FusionEXNet, an innovative and interpretable fused deep-learning model for skin cancer detection. FusionEXNet leverages the strengths of EfficientNetV2S and XceptionNet architectures to extract robust features from dermoscopic images, achieving superior performance compared to individual models. While XceptionNet and EfficientNetV2S attained accuracies of 88.82% and 88.01%, respectively, FusionEXNet surpassed them with an impressive accuracy of 90.83%. To enhance model interpretability, Explainable Artificial Intelligence (XAI) techniques, such as SmoothGrad and Faster Score-CAM, are integrated, providing valuable insights into the decision-making process. The model was trained and evaluated using the extensive HAM10000 dataset, consisting of over 10,000 high-resolution images across seven skin lesion categories. The proposed FusionEXNet model offers a reliable, accurate, and interpretable system for skin cancer detection, contributing to improved patient outcomes and more efficient diagnostic processes. This study underscores the potential of combining advanced CNN architectures with XAI methods to create powerful tools for dermatological diagnostics.

FAULT DETECTION AND CLASSIFICATION USING DENSENET, SURF-BASED ANN, AND INFRARED THERMOGRAPHY

The reliability and efficiency of electric motors are critical in industrial applications, where unexpected faults can lead to costly downtime and safety hazards. Traditional fault detection methods often require extensive manual inspection and may not capture subtle anomalies in motor behavior. Infrared thermography has emerged as a non-invasive technique to detect temperature variations in motor components, which can indicate potential faults. However, the challenge lies in accurately classifying these faults to prevent failures. Current methods lack the precision needed to classify various motor faults accurately and quickly, especially when dealing with complex thermal patterns. The integration of advanced deep learning architectures with feature extraction techniques presents an opportunity to enhance the detection and classification of motor faults. This study proposes a hybrid model combining DenseNet, a deep learning architecture known for its high performance in image analysis, with a Speeded-Up Robust Features (SURF)-based Artificial Neural Network (ANN) for feature extraction and classification. Infrared thermography images of motors were first processed through DenseNet for initial feature extraction. The SURF algorithm further refined these features, which were then classified using ANN. The model was trained and validated on a dataset of infrared thermography images, representing various motor fault conditions, including bearing wear, misalignment, and insulation failure. The proposed model achieved an overall accuracy of 98.7% in detecting and classifying motor faults, outperforming traditional methods by 5.2%. The model also demonstrated high sensitivity (97.8%) and specificity (99.1%) in identifying subtle temperature variations indicative of early-stage faults. These results highlight the effectiveness of the DenseNet and SURF-based ANN approach in enhancing the reliability of motor fault detection using infrared thermography.

Personalized Recommendation Systems Powered By Large Language Models: Integrating Semantic Understanding and User Preferences

This study proposes a novel personalized recommendation system leveraging Large Language Models (LLMs) to integrate semantic understanding with user preferences [1]. The system addresses critical challenges in traditional recommendation approaches by harnessing LLMs' advanced natural language processing capabilities. We introduce a framework combining a fine-tuned Roberta semantic analysis model with a multi-modal user preference extraction mechanism.The LLM component undergoes domain adaptation using Masked Language Modeling on a corpus of 112,000 user reviews from the MyAnimeList dataset, followed by task-specific fine-tuning using contrastive learning. User preferences are modeled through a weighted combination of explicit ratings, review sentiment, and implicit feedback, incorporating temporal dynamics through a time-decay function. Experimental results demonstrate significant improvements over state-of-the-art baselines, including Matrix Factorization, Neural Collaborative Filtering, BERT4Rec, and LightGCN. Our LLM-powered system achieves an 8.6%increase in NDCG@10 and a 10.5% improvement in Mean Reciprocal Rank compared to the best-performing baseline. Ablation studies reveal the synergistic effect of integrating LLM-based semantic understanding with user preference modeling. Case studies highlight the system's ability to recommend long-tail items and provide cross-genre suggestions, showcasing its capacity for nuanced content understanding. Scalability analysis indicates that while the LLM-based approach has higher initial computational costs, its performance scales comparably to other deep learning models for larger datasets. This research contributes to the field by demonstrating the effectiveness of LLMs in enhancing recommendation accuracy and diversity. Future work will explore advanced LLM architectures, multi-modal data integration, and techniques to improve computational efficiency and interpretability of recommendations.

Semantic Segmentation in Large-Size Orthomosaics to Detect the Vegetation Area in Opuntia spp. Crop.

This study focuses on semantic segmentation in crop Opuntia spp. orthomosaics; this is a significant challenge due to the inherent variability in the captured images. Manual measurement of Opuntia spp. vegetation areas can be slow and inefficient, highlighting the need for more advanced and accurate methods. For this reason, we propose to use deep learning techniques to provide a more precise and efficient measurement of the vegetation area. Our research focuses on the unique difficulties posed by segmenting high-resolution images exceeding 2000 pixels, a common problem in generating orthomosaics for agricultural monitoring. The research was carried out on a Opuntia spp. cultivation located in the agricultural region of Tulancingo, Hidalgo, Mexico. The images used in this study were obtained by drones and processed using advanced semantic segmentation architectures, including DeepLabV3+, UNet, and UNet Style Xception. The results offer a comparative analysis of the performance of these architectures in the semantic segmentation of Opuntia spp., thus contributing to the development and improvement of crop analysis techniques based on deep learning. This work sets a precedent for future research applying deep learning techniques in agriculture.

STCPU-Net: advanced U-shaped deep learning architecture based on Swin transformers and capsule neural network for brain tumor segmentation

STCPU-Net: advanced U-shaped deep learning architecture based on Swin transformers and capsule neural network for brain tumor segmentation