Search Results (2,112)

The rising computational and energy demands of artificial intelligence systems urge the exploration of alternative software and hardware solutions that exploit physical effects for computation. According to machine learning theory, a neural network-based computational system must exhibit nonlinearity to effectively model complex patterns and relationships. This requirement has driven extensive research into various nonlinear physical systems to enhance the performance of neural networks. In this paper, we propose and theoretically validate a reservoir-computing system based on a single bubble trapped within a bulk of liquid. By applying an external acoustic pressure wave to both encode input information and excite the complex nonlinear dynamics, we showcase the ability of this single-bubble reservoir-computing system to forecast a Hénon benchmarking time series and undertake classification tasks with high accuracy. Specifically, we demonstrate that a chaotic physical regime of bubble oscillation—where tiny differences in initial conditions lead to wildly different outcomes, making the system unpredictable despite following clear rules, yet still suitable for accurate computations—proves to be the most effective for such tasks. Full article

Chrysanthemum has a long history of medicinal use with rich germplasm resources and extensive cultivation. Traditional chrysanthemum cultivation involves complex patterns and long flowering periods, with the ongoing expansion of planting areas complicating statistical surveys. Currently, reliable, timely, and universally applicable standardized monitoring methods for chrysanthemum cultivation areas remain underdeveloped. This research employed 16 m resolution satellite imagery spanning 2021 to 2023 alongside 2 m resolution data acquired in 2022 to quantify chrysanthemum cultivation extent across Sheyang County, Jiangsu Province, China. After evaluating multiple classifiers, Maximum Likelihood Classification was selected as the optimal method. Subsequently, time-series-based post-classification processing was implemented: initial cultivation information extraction was performed through feature comparison, supervised classification, and temporal analysis. Accuracy validation via Overall Accuracy, Kappa coefficient, Producer’s Accuracy, and User’s Accuracy identified critical issues, followed by targeted refinement of spectrally confused features to obtain precise area estimates. The chrysanthemum cultivation area in 2022 was quantified as 46,950,343 m² for 2 m resolution and 46,332,538 m² for 16 m resolution. Finally, the conversion ratio characteristics between resolutions were analyzed, yielding adjusted results of 38,466,192 m² for 2021 and 47,546,718 m² for 2023, respectively. These outcomes demonstrate strong alignment with local agricultural statistics, confirming method viability for chrysanthemum cultivation area computation. Full article

This work contextualizes the possibility of deriving a unifying artificial intelligence framework by walking in the footsteps of General, Explainable, and Verified Artificial Intelligence (GEVAI): by considering explainability not only at the level of the results produced by a specification but also considering the explicability of the inference process as well as the one related to the data processing step, we can not only ensure human explainability of the process leading to the ultimate results but also mitigate and minimize machine faults leading to incorrect results. This, on the other hand, requires the adoption of automated verification processes beyond system fine-tuning, which are essentially relevant in a more interconnected world. The challenges related to full automation of a data processing pipeline, mostly requiring human-in-the-loop approaches, forces us to tackle the framework from a different perspective: while proposing a preliminary implementation of GEVAI mainly used as an AI test-bed having different state-of-the-art AI algorithms interconnected, we propose two other data processing pipelines, LaSSI and EMeriTAte+DF, being a specific instantiation of GEVAI for solving specific problems (Natural Language Processing, and Multivariate Time Series Classifications). Preliminary results from our ongoing work strengthen the position of the proposed framework by showcasing it as a viable path to improve current state-of-the-art AI algorithms. Full article

Existing approaches to social media sentiment analysis typically focus on static classification, offering limited foresight into how public opinion evolves. This study addresses that gap by introducing the Multi-Feature Sentiment-Driven Forecasting (MFSF) framework, a novel pipeline that enhances sentiment trend prediction by integrating rich contextual information from text. Using state-of-the-art transformer models on the Sentiment140 dataset, our framework extracts three concurrent signals from each tweet: sentiment polarity, aspect-based scores (e.g., ‘price’ and ‘service’), and topic embeddings. These features are aggregated into a daily multivariate time series. We then employ a SARIMAX model to forecast future sentiment, using the extracted aspect and topic data as predictive exogenous variables. Our results, validated on the historical Sentiment140 Twitter dataset, demonstrate the framework’s superior performance. The proposed multivariate model achieved a 26.6% improvement in forecasting accuracy (RMSE) over a traditional univariate ARIMA baseline. The analysis confirmed that conversational aspects like ‘service’ and ‘quality’ are statistically significant predictors of future sentiment. By leveraging the contextual drivers of conversation, the MFSF framework provides a more accurate and interpretable tool for businesses and policymakers to proactively monitor and anticipate shifts in public opinion. Full article

Emotion analysis based on electroencephalogram (EEG) sensors is pivotal for human–machine interaction yet faces key challenges in spatio-temporal feature fusion and cross-band and brain-region integration from multi-channel sensor-derived signals. This paper proposes MB-MSTFNet, a novel framework for EEG emotion recognition. The model constructs a 3D tensor to encode band–space–time correlations of sensor data, explicitly modeling frequency-domain dynamics and spatial distributions of EEG sensors across brain regions. A multi-scale CNN-Inception module extracts hierarchical spatial features via diverse convolutional kernels and pooling operations, capturing localized sensor activations and global brain network interactions. Bi-directional GRUs (BiGRUs) model temporal dependencies in sensor time-series, adept at capturing long-range dynamic patterns. Multi-head self-attention highlights critical time windows and brain regions by assigning adaptive weights to relevant sensor channels, suppressing noise from non-contributory electrodes. Experiments on the DEAP dataset, containing multi-channel EEG sensor recordings, show that MB-MSTFNet achieves 96.80 ± 0.92% valence accuracy, 98.02 ± 0.76% arousal accuracy for binary classification tasks, and 92.85 ± 1.45% accuracy for four-class classification. Ablation studies validate that feature fusion, bidirectional temporal modeling, and multi-scale mechanisms significantly enhance performance by improving feature complementarity. This sensor-driven framework advances affective computing by integrating spatio-temporal dynamics and multi-band interactions of EEG sensor signals, enabling efficient real-time emotion recognition. Full article

Background/Objectives: Radial head arthroplasty is a commonly preferred treatment for complex, unreconstructable radial head fractures. Recent studies have raised the question of whether factors such as bone resorption may be related to failure. This observational, retrospective, multicenter, spontaneous, and non-profit study aims to assess radiological outcomes, focusing on bone resorption around the stem, for radial head replacement using a modular, cementless radial head prosthesis with interconnected porosity. Methods: A series of 42 cases was available for review. Patients underwent radial head arthroplasty using a three-dimensional-printed radial head prosthesis. Patients were eligible for inclusion if they had undergone at least one follow-up between 6 and 15 months post-operatively. A scoring system to detect bone resorption was developed and administered by two independent evaluators. Results: Forty-two patients (14 males, 28 females), with an average age of 59 ± 11 years (range: 39–80 years), were analyzed with a minimum of six months’ and a maximum of 32 months’ follow-up. At follow-up, 50 radiological evaluations were collected, with 29 showing ≤3 mm and 12 showing 3–6 mm resorption around the stem. The average resorption was 3.5 mm ± 2.3. No correlation was found between the extent of resorption and the time of follow-up. The developed scoring system allowed for a high level of correlation between the evaluators’ measurements of bone resorption. Conclusions: Radial head prosthesis with interconnected porosity provided a low stem resorption rate for patients after a radial head fracture at short-to-mid-term follow-up after the definition of a reliable and easy-to-use radiological-based classification approach. (Level of Evidence: Level IV). Full article

The mapping and monitoring of coastal regions are critical to ensure their sustainable use and viability in the long term. Delineation of coastlines is becoming increasingly important in the light of climate change and rising sea levels. However, many coastlines are highly dynamic; therefore, mono-temporal assessments of coastal ecosystems and coastlines are mere snapshots of limited practical value for space-based planning. Understanding of the spatio-temporal dynamics of coastal ecosystem boundaries is important to inform ecosystem management but also for a meaningful delineation of the high-water mark, which is used as a benchmark for coastal spatial planning in South Africa. This research aimed to use hyper-temporal Sentinel-2 imagery to extract ecological zones on the coast of KwaZulu-Natal, South Africa. A total of 613 images, collected between 2019 and 2023, were classified into four distinct coastal ecological zones—vegetation, bare, surf, and water—using a Random Forest model. Across all classifications, the percentage of each of the four classes’ occurrence per pixel over time was determined. This enabled the identification of ecosystem locations, spatially static ecosystem boundaries, and the occurrence of ecosystem boundaries with a more dynamic location over time, such as the non-permanent vegetation zone of the foredune area as well as the intertidal zone. The overall accuracy of the model was 98.13%, while the Kappa coefficient was 0.975, with user’s and producer’s accuracies ranging between 93.02% and 100%. These results indicate that cloud-based analysis of Sentinel-2 time series holds potential not just for delineating coastal ecosystem boundaries, but also for enhancing the understanding of spatio-temporal dynamics between them, to inform meaningful environmental management, spatial planning, and climate adaptation strategies. Full article

In industrial systems, sensors often classify collected time series data for incipient fault diagnosis. However, time series data from sensors during the initial stages of a fault often exhibits minor fluctuation characteristics. Existing time series classification (TSC) methods struggle to achieve high classification accuracy when these minor fluctuations serve as the primary distinguishing feature. This limitation arises because the low-amplitude variations of these fluctuations, compared with trends, lead the classifier to prioritize and learn trend features while ignoring the minor fluctuations crucial for accurate classification. To address this challenge, this paper proposes a novel graph-based time series classification framework, termed MFSI-TSC. MFSI-TSC first extracts the trend component of the raw time series. Subsequently, both the trend series and the raw series are represented as graphs by extracting the “visible relationship” of the series. By performing a subtraction operation between these graphs, the framework isolates the differential information arising from the minor fluctuations. The subtracted graph effectively captures minor fluctuations by highlighting topological variations, thereby making them more distinguishable. Furthermore, the framework incorporates optimizations to reduce computational complexity, facilitating its deployment in resource-constrained sensor systems. Finally, empirical evaluation of MFSI-TSC on both real-world and publicly available datasets demonstrates its effectiveness. Compared with ten benchmark methods, MFSI-TSC exhibits both high accuracy and computational efficiency, making it more suitable for deployment in sensor systems to complete incipient fault detection tasks. Full article

This study presents a computational analysis of industry-specific advertising message strategies through the theoretical lens of the FCB (Foote, Cone & Belding) grid framework. Leveraging the AiSAC (AI Analysis System for Ad Creation) system developed by the Korea Broadcast Advertising Corporation (KOBACO), we analyzed 27,000 Korean advertisements across five major industries using advanced machine learning techniques. Through Latent Dirichlet Allocation topic modeling with a coherence score of 0.78, we identified five distinct message strategies: emotional appeal, product features, visual techniques, setting and objects, and entertainment and promotion. Our computational analysis revealed that each industry exhibits a unique “message strategy fingerprint” that significantly discriminates between categories, with discriminant analysis achieving 62.7% classification accuracy. Time-series analysis using recurrent neural networks demonstrated a significant evolution in strategy preferences, with emotional appeal increasing by 44.3% over the study period (2015–2024). By mapping these empirical findings onto the FCB grid, the present study validated that industry positioning within the grid’s quadrants aligns with theoretical expectations: high-involvement/think (IT and Telecom), high-involvement/feel (Public Institutions), low-involvement/think (Food and Household Goods), and low-involvement/feel (Services). This study contributes to media science by demonstrating how computational methods can empirically validate the established theoretical frameworks in advertising, providing a data-driven approach to understanding message strategy patterns across industries. Full article

As the demand for more accurate crime prediction and risk assessment grows, researchers have been developing smarter models that blend statistical methods with machine learning. This study compares a hybrid ARIMA-ANN model with traditional classification techniques to see which best forecast monthly crime risk levels in Galați County, Romania. The analysis is based on a newly compiled dataset of 132 monthly observations from January 2014 to December 2024, which combines a broad array of social, economic, and environmental data points. The main variable, ‘Crime risk’, is based on normalized counts of offenses per capita and divided into five balanced levels: very low, low, moderate, high, and very high. The hybrid ARIMA-ANN model merges the strengths of statistical time series analysis with the flexible learning ability of artificial neural networks. Performance is evaluated against multinomial logistic regression, decision trees, random forests, and support vector machines. Overall, the results show that an ARIMA-ANN model consistently outperforms traditional methods, especially in recognizing patterns over time, seasonal trends, and complex nonlinear relationships in crime data. This study not only sets a new benchmark for crime analytics in Romania but also offers a flexible, scalable framework for classifying crime risk levels across different regions. Full article

Federated learning (FL) enables multiple participants to collaboratively train models while efficiently mitigating the issue of data silos. However, large-scale heterogeneous data distributions result in inconsistent client objectives and catastrophic forgetting, leading to model bias and slow convergence. To address the challenges under non-independent and identically distributed (non-IID) data, we propose DPAO-PFL, a Dynamic Parameter-Aware Optimization framework that leverages continual learning principles to improve Personalized Federated Learning under non-IID conditions. We decomposed the parameters into two components: local personalized parameters tailored to client characteristics, and global shared parameters that capture the accumulated marginal effects of parameter updates over historical rounds. Specifically, we leverage the Fisher information matrix to estimate parameter importance online, integrate the path sensitivity scores within a time-series sliding window to construct a dynamic regularization term, and adaptively adjust the constraint strength to mitigate the conflict overall tasks. We evaluate the effectiveness of DPAO-PFL through extensive experiments on several benchmarks under IID and non-IID data distributions. Comprehensive experimental results indicate that DPAO-PFL outperforms baselines with improvements from $5.41\%$ to $30.42\%$ in average classification accuracy. By decoupling model parameters and incorporating an adaptive regularization mechanism, DPAO-PFL effectively balances generalization and personalization. Furthermore, DPAO-PFL exhibits superior performance in convergence and collaborative optimization compared to state-of-the-art FL methods. Full article

Ergonomic risk classification during manual lifting tasks is crucial for the prevention of workplace injuries. This study addresses the challenge of classifying lifting task risk levels (low, medium, and high risk, labeled as 0, 1, and 2) using multi-modal time-series data comprising raw facial landmarks and bio-signals (electrocardiography [ECG] and electrodermal activity [EDA]). Classifying such data presents inherent challenges due to multi-source information, temporal dynamics, and class imbalance. To overcome these challenges, this paper proposes a Multi-Adaptive Functional Neural Network (Multi-AdaFNN), a novel method that integrates functional data analysis with deep learning techniques. The proposed model introduces a novel adaptive basis layer composed of micro-networks tailored to each individual time-series feature, enabling end-to-end learning of discriminative temporal patterns directly from raw data. The Multi-AdaFNN approach was evaluated across five distinct dataset configurations: (1) facial landmarks only, (2) bio-signals only, (3) full fusion of all available features, (4) a reduced-dimensionality set of 12 selected facial landmark trajectories, and (5) the same reduced set combined with bio-signals. Performance was rigorously assessed using 100 independent stratified splits (70% training and 30% testing) and optimized via a weighted cross-entropy loss function to manage class imbalance effectively. The results demonstrated that the integrated approach, fusing facial landmarks and bio-signals, achieved the highest classification accuracy and robustness. Furthermore, the adaptive basis functions revealed specific phases within lifting tasks critical for risk prediction. These findings underscore the efficacy and transparency of the Multi-AdaFNN framework for multi-modal ergonomic risk assessment, highlighting its potential for real-time monitoring and proactive injury prevention in industrial environments. Full article

The spatial distribution of ephemeral and perennial dryland plant species is increasingly modified and restricted by ever-changing climates and development expansion. At the interface of biodiversity conservation and developmental planning in desert landscapes is the growing need for adaptable tools in identifying and monitoring these ecologically fragile plant assemblages, habitats, and, often, heritage sites. This study evaluates usage of Sentinel-2 time series composite imagery to discriminate vegetation assemblages in a hyper-arid landscape. Spatial predictor spaces were compared to classify different vegetation communities: spectral components (PCs), vegetation indices (VIs), and their combination. Further, the uncertainty in discriminating field-verified vegetation assemblages is assessed using Shannon entropy and intensity analysis. Lastly, the intensity analysis helped to decipher and quantify class transitions between maps from different spatial predictors. We mapped plant assemblages in 2022 from combined PCs and VIs at an overall accuracy of 82.71% (95% CI: 81.08, 84.28). A high overall accuracy did not directly translate to high class prediction probabilities. Prediction by spectral components, with comparably lower accuracy (80.32, 95% CI: 78.60, 81.96), showed lower class uncertainty. Class disagreement or transition between classification models was mainly contributed by class exchange (a component of spatial allocation) and less so from quantity disagreement. Different artefacts of vegetation classes are associated with the predictor space—spectral components versus vegetation indices. This study contributes insights into using feature extraction (VIs) versus feature selection (PCs) for pixel-based classification of plant assemblages. Emphasising the ecologically sensitive vegetation in desert landscapes, the study contributes uncertainty considerations in translating optical satellite imagery to vegetation maps of arid landscapes. These are perceived to inform and support vegetation map creation and interpretation for operational management and conservation of plant biodiversity and habitats in such landscapes. Full article

Image analysis is a valuable approach in a wide array of environmental applications. Mapping land cover categories depicted from satellite images enables the monitoring of landscape dynamics. Such a technique plays a key role for land management and predictive ecosystem modelling. Satellite-based mapping of environmental dynamics enables us to define factors that trigger these processes and are crucial for our understanding of Earth system processes. In this study, a reclassification scheme of image analysis was developed for mapping the adjusted categorisation of land cover types using multispectral remote sensing datasets and Geographic Resources Analysis Support System (GRASS) Geographic Information System (GIS) software. The data included four Landsat 8–9 satellite images on 2015, 2019, 2021 and 2023. The sequence of time series was used to determine land cover dynamics. The classification scheme consisting of 17 initial land cover classes was employed by logical workflow to extract 10 key land cover types of the coastal areas of Bab-el-Mandeb Strait, southern Red Sea. Special attention is placed to identify changes in the land categories regarding the thermal saline lake, Lake Assal, with fluctuating salinity and water levels. The methodology included the use of machine learning (ML) image analysis GRASS GIS modules ‘r.reclass’ for the reclassification of a raster map based on category values. Other modules included ‘r.random’, ‘r.learn.train’ and ‘r.learn.predict’ for gradient boosting ML classifier and ‘i.cluster’ and ‘i.maxlik’ for clustering and maximum-likelihood discriminant analysis. To reveal changes in the land cover categories around the Lake of Assal, this study uses ML and reclassification methods for image analysis. Auxiliary modules included ‘i.group’, ‘r.import’ and other GRASS GIS scripting techniques applied to Landsat image processing and for the identification of land cover variables. The results of image processing demonstrated annual fluctuations in the landscapes around the saline lake and changes in semi-arid and desert land cover types over Djibouti. The increase in the extent of semi-desert areas and the decrease in natural vegetation proved the processes of desertification of the arid environment in Djibouti caused by climate effects. The developed land cover maps provided information for assessing spatial–temporal changes in Djibouti. The proposed ML-based methodology using GRASS GIS can be employed for integrating techniques of image analysis for land management in other arid regions of Africa. Full article

An underground utility tunnel (UUT) is essential for the efficient use of urban underground space. However, current maintenance systems rely on patrol personnel and professional equipment. This study explores industrial detection methods for identifying and monitoring natural gas leaks in UUTs. Via infrared thermal imaging gas experiments, data were acquired and a dataset established. To address the low-resolution problem of existing imaging devices, video super-resolution (VSR) was used to improve the data quality. Based on a convolutional recurrent neural network (CRNN), the image features at each moment were extracted, and the time series data were modeled to realize the risk-level classification mechanism based on the automatic classification of the leakage rate. The experimental results show that when the sliding window size was set to 10 frames, the classification accuracy of the CRNN was the highest, which could reach 0.98. This method improves early warning precision and response efficiency, offering practical technical support for UUT maintenance management. Full article