Accurate enzyme specificity constant prediction with iESC.

In the c formation and automation, efficient ai.d accurate text recognition plays a crucial role in bridging the gap between human writing and comp riting and compu, interpretation. Our project, the Handwritten Character Recognition (HCR) System using leural Networks, addresses this need by establishing an intelligent model capable of identifying and classifying handwi racter with high precision. The system utilizes deep learning techniques, specifically Convolutional Neurai Networks (CNNs), to analyze handwritten input images and cotvert them into machine-readable text. Built on a scalab t neural network architecture, this system employs advanced prepro- cessing methods su nage normalization, noise reduction, and segmentation to ensure accurate recognition. The crained using a large set of handwritien samples, enabling the model to learn diverse writing styles and patterns. The The trained model then predicts the corresponding characters with improved accuracy through multiple feature extre tion and classification layers.

ABSTRACT Housing prices incorporate a variety of factors, such as specifications of the house, the neighbourhood environment, and the personal preferences of home buyers. The main purpose of this study is to explore the relationship between images and text as unstructured big data and housing prices. For this purpose, deep learning-based semantic segmentation and text mining-based sentiment analyses were used to measure the visual characteristics of streets and residents’ sentiments about and awareness of the environment. The main results of the analysis are as follows. First, streetscape characteristics (greenness, openness, enclosure, and walkability) are associated with variations in apartment prices. Second, a nonlinear relationship was found between openness, enclosure, and apartment prices. Specifically, apartment prices decreased as openness increased but increased after a certain level, while enclosure exhibited the opposite pattern. Third, apartment complex reviews and apartment prices had a significant, nonlinear relationship. As the sentiment of apartment complex reviews became more positive, apartment prices decreased and then increased after a certain level. Fourth, residents’ perceptions of the apartment complex brand had a significant relationship with marketability and apartment prices. Specifically, the higher was the brand awareness, represented by the construction company's contract ranking, the higher was the apartment price. These findings suggest that both the visual characteristics of streetscapes and residents’ perceptions of their living environment contribute to explaining variations in housing prices, with nonlinear effects observed across multiple factors.

Aerospace engines, as critical components in the aviation and aerospace industries, require continuous and accurate fault diagnosis to ensure operational safety and prevent catastrophic failures. While deep learning techniques have been extensively studied in this context, they typically output logits or confidence scores, necessitating postprocessing to obtain actionable insights. Furthermore, the potential of large-scale audio models for this task remains largely untapped. To address these limitations, this article proposes AeroGPT, a novel framework that transfers knowledge from the general audio domain to aero-engine bearing fault diagnosis. AeroGPT leverages a large-scale audio model and incorporates vibration signal alignment (VSA) to adapt general audio knowledge to domain-specific vibration patterns, along with generative fault classification (GFC) to directly generate interpretable fault labels. This approach eliminates the need for label postprocessing and supports interactive, interpretable, and actionable fault diagnosis, thereby enhancing industrial applicability. Through comprehensive experimental validation on two aero-engine bearing datasets, AeroGPT achieves 98.94% accuracy on the Dynamic and Identification Research Group (DIRG) dataset and 100% accuracy on Harbin Institute of Technology (HIT) bearing dataset, outperforming representative deep learning approaches. Qualitative analysis and further discussion also demonstrate its potential for interactive diagnosis and real-world deployment, highlighting the promise of large-scale audio models to advance fault diagnosis in aerospace applications.

Advanced attention mechanisms considerably improve crop classification using time-series and frequency information. Self-attention reigns supreme at identifying detailed temporal patterns and pointing out key features within remote sensing data. Tanh-activated self-attention achieved the highest accuracy (88.89%), over multiplicative attention (85.67%), soft attention (82.98%) and global attention (82.12%). Focusing on key time-based and frequency-based patterns, these techniques helped in improving the model's ability to accurately differentiate between crop types. The study introduces a new method for accurate crop classification; this method uses vegetation indices along with attention mechanisms. Agricultural monitoring faces challenges. These include temporal changes, complex spectral data and variable ecological conditions. This method incorporates vegetation index data and attention-based deep learning to address them. This framework investigates the interplay of vegetation indices along with attention mechanisms across multiple ecological conditions as well as plant growth stages, using advanced techniques such as data resampling, feature engineering and machine learning.

Progress in chemistry has been driven by the streamlining of inverse problem-solving methods. In the history of chemistry, several revolutionary technologies have led to leaps forward: the establishment of atomistic theory in the 19th century, structural analysis by spectroscopy in the 20th century, and the development of simulation by theoretical chemistry. Currently, chemistry is about to make a significant leap forward by integrating generative artificial intelligence (AI). In 2016, deep learning techniques were introduced in this domain, leading to explosive development. This paper reviews the development path, including traditional models such as variational autoencoders and more up-to-date models such as large language models and diffusion models. We also discuss how AI can have a real impact on chemistry, including the possibilities and problems associated with synthesizing AI-generated molecules.

Abstract - The identification, segmentation, and extraction of tumor-affected regions from Magnetic Resonance Imaging (MRI) scans are important tasks in medical diagnosis. These processes are usually performed by radiologists or medical experts and require significant time and experience. Image processing techniques help in visualizing the anatomical structures of human organs more effectively, but detecting abnormal structures in the brain using basic imaging methods is still difficult. In this study, a Fully Automatic Heterogeneous Segmentation using Support Vector Machine (FAHS-SVM) method is proposed for brain tumor segmentation using deep learning techniques. The approach introduces a fully automated algorithm that separates the cerebral venous system in MRI images by utilizing structural, morphological, and relaxometry information. The segmentation process ensures a high level of consistency between the brain anatomy and the surrounding tissues. Extreme Learning Machine (ELM), which contains one or more hidden layers, is applied as a learning algorithm and is commonly used for tasks such as regression and classification. In this work, a probabilistic neural network classifier is used to train and evaluate the detection accuracy of tumors in brain MRI images. Experimental results demonstrate that the proposed system can effectively distinguish between normal and abnormal brain tissues with an accuracy of approximately 98.51%, showing the effectiveness of the proposed method.

Mental health monitoring through emotion recognition plays an important role in early intervention and personalized healthcare systems. Traditional EEG-based emotion recognition approaches have encountered significant limitations, including heavy reliance on manual feature engineering, poor generalization across datasets, and computational complexity that restricts real-world deployment. This study introduces a novel hybrid dual-branch deep learning architecture that integrates temporal and spectral feature extraction for robust EEG-based emotion recognition, while minimizing preprocessing requirements. The proposed framework integrates a Long Short-Term Memory (LSTM) network to capture temporal dependencies directly from raw EEG signals, while concurrently leveraging Convolutional Neural Networks (CNNs) to extract spatial features from Mel-Frequency Cepstral Coefficients (MFCC) representations. This architecture further incorporates innovative cross-modality enhancement mechanisms, such as inverse MFCC computation and LSTM-to-MFCC projection, which facilitate bidirectional feature learning between temporal and spectral domains. Subsequently, feature fusion is achieved through element-wise multiplication and concatenation, and the integrated representations are classified using an Artificial Neural Network (ANN). The evaluation has been conducted on three benchmark datasets: Brainwave EEG, WESAD, and SWELL, achieving remarkable performance with accuracies of 96.49%, 99.99%, and 99.99%, respectively. The model has attained perfect precision, recall, and F1-scores on the WESAD and SWELL datasets, setting new state-of-the-art standards. Comparative analysis has demonstrated the method's superiority over existing machine learning and deep learning techniques, while preserving computational efficiency that supports real-time applications. These results have confirmed the framework's strong potential for practical use in mental health monitoring and affective computing.The source code of this work is available at:https://github.com/DebamSahaCS/Dual-Branch-DL-Framework.

Abstract. Climate change and sea-level rise (SLR) pose escalating threats to coastal cities, intensifying the need for efficient and accurate methods to predict potential flood hazards. Traditional physics-based hydrodynamic simulators, although precise, are computationally expensive and impractical for city-scale coastal planning applications. Deep Learning (DL) techniques offer promising alternatives, however, they are often constrained by challenges such as data scarcity and high-dimensional output requirements. Leveraging a recently proposed vision-based, low-resource DL framework, we develop a novel, lightweight Convolutional Neural Network (CNN)-based model designed to predict coastal flooding under variable SLR projections and shoreline adaptation scenarios. Furthermore, we demonstrate the ability of the model to generalize across diverse geographical contexts by utilizing datasets from two distinct regions: Abu Dhabi and San Francisco. Our findings demonstrate that the proposed model significantly outperforms state-of-the-art methods, reducing the mean absolute error (MAE) in predicted flood depth maps on average by nearly 20 %. These results highlight the potential of our approach to serve as a scalable and practical tool for coastal flood management, empowering decision-makers to develop effective mitigation strategies in response to the growing impacts of climate change. Project Page: https://caspiannet.github.io/ (last access: 22 January 2026).

Use of synthetic data, a novel paradigm for immunopathology.

Purpose This study aims to explore the temporal evolution of user-generated popular tags in Flickr, a collaborative image tagging system, through the lens of facet classification. It aims to uncover how tagging behaviors reflect shifts in technology, culture and individual cognition and how deep learning techniques can enhance our understanding of the semantic structures within folksonomies. Design/methodology/approach Popular tags from three benchmark years (2006, 2010 and 2015) were collected and manually categorized using a faceted classification scheme rooted in Ranganathan’s model. To augment this analysis, we applied autoencoder-based deep learning models to extract latent semantic representations and pretrained word embeddings to measure semantic similarity. This hybrid approach enabled both qualitative categorization and quantitative analysis of temporal-semantic patterns in tag usage. Findings The study found that over 80% of the most popular tags were consistently associated with spatial, personality and material facets. Temporal analysis revealed a marked shift from time-based tags to more self-expressive, personality-oriented tags, reflecting users’ growing inclination toward individualism and identity expression. Location-based tags remained persistent across all years, suggesting the global, place-based nature of user engagement. Deep learning – enhanced analysis revealed semantic groupings and confirmed that tagging behavior evolves in alignment with technological and cultural developments. Originality/value This study offers a novel integration of traditional facet analysis (colon classification) and deep learning to model the dynamics of user-generated metadata on social image platforms. It advances understanding of how folksonomies, when enriched by neural representation learning, can serve as both mirrors of cultural shifts and tools for more adaptive, user-centered metadata systems. The approach contributes methodologically to digital classification research and provides insights into the cognitive and social factors shaping online tagging behavior.

AI-driven fault detection and classification in photovoltaic systems using deep learning techniques.

A deep reinforcement learning algorithm for optimizing safety and efficiency of traffic signals using traffic conflict technique and artificial intelligence-based video analytics.

Concordance matrices play a crucial role in input–output analysis, for translating between databases expressed in different sector classifications, or translating many misaligned databases into a common format. These matrices are critical tools in enabling the utilisation of all possible primary data sources for compiling input–output tables, even if those data sources are completely misaligned. Until this date, concordance matrices are constructed manually by interpreting pairwise sector labels, resulting in an often labour-intensive process. In this work, we use artificial intelligence (AI) approaches for the first time to estimate concordance matrices for input–output analysis, offering to significantly reduce the time and labour involved in primary data processing. We show that, when applying deep learning techniques to textual sector labels, AI algorithms are able to grasp intricate linguistic relationships and capture semantic nuances, thus bridging the gap between human language and numerical binary relationships. We use a range of performance evaluation measures and demonstrate the ability to predict a wide range of concordance matrices with up to 85% accuracy.

Estimating conditional extreme quantiles is essential for assessing tail risks in complex systems, with broad applications in finance, climate science, engineering, and beyond. While classical extreme value theory provides a foundational framework, recent advances, particularly semiparametric and nonparametric methods, including approaches based on quantile regression, machine learning, and deep learning, have greatly enriched the methodological landscape. This modern review synthesizes these developments, covering traditional likelihood-based methods, semiparametric approaches, and tree-based and deep learning techniques, including higher-order refinements.

Retraction Note to: Highway traffic congestion detection and evaluation based on deep learning techniques

Despite Pattani's rich biodiversity, there is a significant lack of digital datasets and automated tools to preserve local ethnomedicinal knowledge. Traditionally, herbal medicine plays a significant role in healthcare systems, especially in regions with rich biodiversity, such as Pattani, Thailand. However, the manual identification of medicinal herbs is often labor-intensive and disposed to inaccuracies. This study introduces an automated system for identifying Pattani's local medicinal herbs (Etlingera elatior, Euphorbia hirta, and Leucas aspera) using Deep Learning methods, specifically Convolutional Neural Networks (CNNs). The system uses image processing techniques to improve the accuracy and confidence in herb identification. The system uses a Convolutional Neural Network (CNN) architecture comprising feature-extraction layers with ReLU activation and max pooling, followed by a fully connected softmax classifier. Data augmentation techniques were employed to enhance model generalization on the collected dataset. It additionally protects traditional knowledge through scientific validation. Collecting and pre-processing a dataset of 600 images and using CNNs yielded an overall test accuracy of 97%. Such performance reinforces the system's capabilities in healthcare for traditional practitioners and pharmaceutical researchers who need precise herb identification. Integrating technology and local knowledge to use and conserve Pattani's medicinal plants and to harness local medicinal plant knowledge is a crucial step this research takes.

With the rapid growth of global maritime trade, the efficient and safe management of maritime traffic has become increasingly critical. This study proposes a comprehensive framework for ship trajectory prediction and maritime traffic congestion identification based on Automatic Identification System (AIS) data. We integrate spatiotemporal analysis with deep learning techniques, specifically combining Graph Convolutional Networks (GCN) and Gated Recurrent Units (GRU) to form a Temporal Graph Convolutional Network (T-GCN) model. This model effectively captures both spatial dependencies among ships and temporal dynamics in traffic flow. Furthermore, we introduce a congestion measurement indicator based on the Speed Performance Index (SPI) to quantify and identify congestion levels in maritime routes. The proposed method not only enhances the accuracy of ship trajectory prediction but also enables proactive congestion warnings, contributing to improved maritime safety and operational efficiency. Experimental results demonstrate the effectiveness of our approach in real-world scenarios.

Sign language makes a significant contribution to enabling communication for speech-impaired individuals across the globe. It needs a physical interpreter to facilitate communication; however, these interpreters are limited in number and may not always be available when needed. Deep learning techniques can help address this challenge by creating a virtual interpreter. However, a few notable challenges such as occlusion, external lighting variations, and background subtraction are to be handled in the recognition system. Gestures performed in Indian Sign Language (ISL) involve movements of a single hand or both hands to convey signs for communication. This article proposes a Static Gesture Categorization and Interpretation of Indian Sign Language (SGCIISLang) model that employs an optimized Gated Recurrent Unit (GRU) architecture. For the experiment, we developed our own dataset titled “Static gestures of Indian Sign Language (ISL) for English Alphabet, Hindi Vowels and Numerals” which has already been published on Mendeley Data (https://data.mendeley.com/datasets/7tsw22y96w/1). The MediaPipe library is used for feature extraction, and the results are integrated into our model to classify sign motions. The region of interest is extracted using three MediaPipe approaches: holistic, holistic without pose, and holistic without face. We analyze and compare four models: Long Short Term Memory-Convolutional Neural Network (LSTM-CNN), Convolutional Neural Network (CNN), Recurrent Neural Network (RNN) and Gated Recurrent Unit (GRU) based SGCIISLang model. Based on the experimental results, the GRU-based SGCIISLang model achieves greater efficiency, faster processing and rapid convergence compared to CNN, LSTM-CNN and RNN models. The prediction precision achieved is 97.96% with loss, recall, mean square error (MSE) and F1 score values of 0.0837, 0.9742, 2.791 and 0.977, respectively. Our approach outperforms and addresses challenges. The sample prototype of our proposed approach is available on GitHub (https://github.com/AnimeshSingh777/Sample-Prototype-for-Indian-Sign-Language-Static-Gesture-Recognition-System).

A fusion of social media context for efficient multimodal sentiment analysis using deep learning techniques