Explore the vast range of titles available.

Audio-visual speech recognition (AVSR) is one of the most promising solutions for reliable speech recognition, particularly when audio is corrupted by noise. Additional visual information can be used for both automatic lip-reading and gesture recognition. Hand gestures are a form of non-verbal communication and can be used as a very important part of modern human–computer interaction systems. Currently, audio and video modalities are easily accessible by sensors of mobile devices. However, there is no out-of-the-box solution for automatic audio-visual speech and gesture recognition. This study introduces two deep neural network-based model architectures: one for AVSR and one for gesture recognition. The main novelty regarding audio-visual speech recognition lies in fine-tuning strategies for both visual and acoustic features and in the proposed end-to-end model, which considers three modality fusion approaches: prediction-level, feature-level, and model-level. The main novelty in gesture recognition lies in a unique set of spatio-temporal features, including those that consider lip articulation information. As there are no available datasets for the combined task, we evaluated our methods on two different large-scale corpora—LRW and AUTSL—and outperformed existing methods on both audio-visual speech recognition and gesture recognition tasks. We achieved AVSR accuracy for the LRW dataset equal to 98.76% and gesture recognition rate for the AUTSL dataset equal to 98.56%. The results obtained demonstrate not only the high performance of the proposed methodology, but also the fundamental possibility of recognizing audio-visual speech and gestures by sensors of mobile devices.

Analysis of microarray data is a highly challenging problem due to the inherent complexity in the nature of the data associated with higher dimensionality, smaller sample size, imbalanced number of classes, noisy data-structure, and higher variance of feature values. This has led to lesser classification accuracy and over-fitting problem. In this work, the authors aimed to develop a deep feedforward method to classify the given microarray cancer data into a set of classes for subsequent diagnosis purposes. They have used a 7-layer deep neural network architecture having various parameters for each dataset. The small sample size and dimensionality problems are addressed by considering a well-known dimensionality reduction technique namely principal component analysis. The feature values are scaled using the Min–Max approach and the proposed approach is validated on eight standard microarray cancer datasets. To measure the loss, a binary cross-entropy is used and adaptive moment estimation is considered for optimisation. The performance of the proposed approach is evaluated using classification accuracy, precision, recall, f-measure, log-loss, receiver operating characteristic curve, and confusion matrix. A comparative analysis with state-of-the-art methods is carried out and the performance of the proposed approach exhibit better performance than many of the existing methods.

The security of Internet of Medical Things (IoMT) devices is crucial for ensuring the integrity and reliability of patients’ medical data. These devices, operating over TCP and ICMP protocols, are highly susceptible to cyberattacks. While machine learning models can detect these attacks with acceptable accuracy, their operational mechanisms remain unclear, leaving the decision-making process of the models undefined. Moreover, the accuracy and training time of machine learning models is more questionable when the datasets has large number of sparse features and class imbalances. This study introduces an interpretable feature selection technique designed to enhance intrusion detection in IoMT by reducing redundant features and improving model efficiency. Random Forest-based explainable AI model provides transparency in attack classification and better decision-making. The simulated results employing the CICIoMT2024 dataset demonstrate that the proposed method significantly improves detection performance, with the Random Forest model achieving 99% accuracy, outperforming XGBoost (98%), Decision Tree (97%), and Support Vector (98%), while ensuring explainability through SHAP-based feature analysis. Thus, the simulation outcomes reveal the key contributing factors for various cyberattacks on IoMT, facilitating enhanced security measures and real-time monitoring. The proposed approach boosts detection accuracy and interpretability, making it highly suitable for real-world IoMT security applications.

In this paper, we develop a method for approximating multiple integrals. The domain of integration Ω is assumed to be a non-rectangular compact of ℝⁿ. The main idea is the dimensionality reduction procedure based on the use of parametric α-dense curves ℓα(t). First, the region whose measure represents the value of the integral, is densified using new results, by a certain α-dense curve of finite length. The multiple integral of a positive continuous function f over Ω is approximated by a unique single integral corresponding to ℓα(t). Some numerical examples are given.

The Brunelleschi’s Dome is one of the most iconic symbols of the Renaissance and is among the largest masonry domes ever constructed. Since the late 17th century, first masonry cracks appeared on the Dome, giving the start to a monitoring activity. In modern times, since 1988 a monitoring system comprised of 166 electronic sensors, including deformometers and thermometers, has been in operation, providing a valuable source of real-time data on the monument’s health status. With the deformometers taking measurements at least four times per day, a vast amount of data is now available to explore the potential of the latest Artificial Intelligence and Machine Learning techniques in the field of historical-architectural heritage conservation. The objective of this contribution is twofold. Firstly, for the first time ever, we aim to unveil the overall structural behaviour of the Dome as a whole, as well as that of its specific sections (known as webs). We achieve this by evaluating the effectiveness of certain dimensionality reduction techniques on the extensive daily detections generated by the monitoring system, while also accounting for fluctuations in temperature over time. Secondly, we estimate a number of recurrent and convolutional neural network models to verify their capability for medium- and long-term prediction of the structural evolution of the Dome. We believe this contribution is an important step forward in the protection and preservation of historical buildings, showing the utility of machine learning in a context in which these are still little used.

Accurate traffic classification plays an important role in efficient utilization of network resources, quality of service, and overall management of the network. The identification of virtual private network (VPN) traffic, in particular, is important since it allows distinguishing between encrypted and non-encrypted traffic by VPN service, which is critical for security monitoring, traffic shaping, and the detection of possible misuse of network resources. VPNs are secure, encrypted connections over an insecure network with predetermined protocols; hence, through traditional methods, it is quite challenging to recognize the traffic pattern. This work introduces the time constrained classification (TCC) model, which use a decision tree classification algorithm with autoencoder dimensionality reduction to extract the key features from encrypted VPN traffic. The TCC model accurately classify VPN traffic from non-VPN traffic without degrading performance and limited amount of time. This approach optimizes the classification time for both binary and multi-class VPN and non-VPN traffic. Experimental results show that the decision tree-based autoencoder model achieves a recall score of 0.993 for multi-class classification in 1.8 s on the UNB ISCX VPN-nonVPN dataset (ISCXVPN2016), outperforming state-of-the-art methods while significantly reducing classification time.

How to effectively obtain species‐related low‐dimensional data from massive environmental variables has become an urgent problem for species distribution models (SDMs). In this study, we will explore whether dimensionality reduction on environmental variables can improve the predictive performance of SDMs. We first used two linear (i.e., principal component analysis (PCA) and independent components analysis) and two nonlinear (i.e., kernel principal component analysis (KPCA) and uniform manifold approximation and projection) dimensionality reduction techniques (DRTs) to reduce the dimensionality of high‐dimensional environmental data. Then, we established five SDMs based on the environmental variables of dimensionality reduction for 23 real plant species and nine virtual species, and compared the predictive performance of those with the SDMs based on the selected environmental variables through Pearson's correlation coefficient (PCC). In addition, we studied the effects of DRTs, model complexity, and sample size on the predictive performance of SDMs. The predictive performance of SDMs under DRTs other than KPCA is better than using PCC. And the predictive performance of SDMs using linear DRTs is better than using nonlinear DRTs. In addition, using DRTs to deal with environmental variables has no less impact on the predictive performance of SDMs than model complexity and sample size. When the model complexity is at the complex level, PCA can improve the predictive performance of SDMs the most by 2.55% compared with PCC. At the middle level of sample size, the PCA improved the predictive performance of SDMs by 2.68% compared with the PCC. Our study demonstrates that DRTs have a significant effect on the predictive performance of SDMs. Specifically, linear DRTs, especially PCA, are more effective at improving model predictive performance under relatively complex model complexity or large sample sizes. Dimensionality reduction techniques (DRTs) can effectively improve the predictive performance of species distribution models by reducing the dimensionality of environmental variables. Specifically, linear DRTs (especially principal component analysis, or PCA) were found to be more effective in improving model performance under relatively complex model complexity or large sample sizes.

Coastal ecosystems experience multiple anthropogenic and climate change pressures. To monitor the variability of the benthic habitats in shallow waters, the implementation of effective strategies is required to support coastal planning. In this context, high-resolution remote sensing data can be of fundamental importance to generate precise seabed maps in coastal shallow water areas. In this work, satellite and airborne multispectral and hyperspectral imagery were used to map benthic habitats in a complex ecosystem. In it, submerged green aquatic vegetation meadows have low density, are located at depths up to 20 m, and the sea surface is regularly affected by persistent local winds. A robust mapping methodology has been identified after a comprehensive analysis of different corrections, feature extraction, and classification approaches. In particular, atmospheric, sunglint, and water column corrections were tested. In addition, to increase the mapping accuracy, we assessed the use of derived information from rotation transforms, texture parameters, and abundance maps produced by linear unmixing algorithms. Finally, maximum likelihood (ML), spectral angle mapper (SAM), and support vector machine (SVM) classification algorithms were considered at the pixel and object levels. In summary, a complete processing methodology was implemented, and results demonstrate the better performance of SVM but the higher robustness of ML to the nature of information and the number of bands considered. Hyperspectral data increases the overall accuracy with respect to the multispectral bands (4.7% for ML and 9.5% for SVM) but the inclusion of additional features, in general, did not significantly improve the seabed map quality.

Recent advancements in dimensionality reduction techniques have significantly contributed to the field of active infrared thermography (AIRT) for defect detection, aiding in data processing and feature extraction. Among these techniques, principal component thermography (PCT) and deep autoencoder thermography (DAT) are particularly notable. PCT is based on conventional linear multivariate analysis, while DAT leverages deep learning paradigms to better handle nonlinearity. These methods consolidate defect information from multiple thermograms into a concise set of feature images, enhancing the visibility of subsurface material defects. However, these feature images often suffer from disturbances, particularly non-uniform backgrounds caused by uneven heating in AIRT experiments. Such interferences can obscure defect information, necessitating further post-processing. In our research, we explore the efficacy of Adaptive Iteratively Reweighted Penalized Least Squares (AIR-PELS) as a refinement technique for PCT and DAT, focusing on background suppression. The adaptive iterative weighting with PELS smoothing effectively reduces noise and removes background disturbances. Case studies involving carbon fiber-reinforced polymer samples with inherent defects demonstrate the effectiveness of this post-processing approach.

The aim of this study is to investigate and demonstrate the role of correction functions and optimisation-based refinement algorithms in enhancing the performance of machine learning models, particularly in predictive anomaly detection tasks applied in industrial environments. The performance of machine learning models is highly dependent on the quality of data preprocessing, model architecture, and post-processing methodology. In many practical applications—particularly in time-series forecasting and anomaly detection—the conventional training pipeline alone is insufficient, because model uncertainty, structural bias and the handling of rare events require specialised post hoc calibration and refinement mechanisms. This study provides a systematic overview of the role of correction functions (e.g., Principal Component Analysis (PCA), Squared Prediction Error (SPE)/Q-statistics, Hotelling’s T2, Bayesian calibration) and adaptive improvement algorithms (e.g., Genetic Algorithms (GA), Particle Swarm Optimisation (PSO), Simulated Annealing (SA), Gaussian Mixture Model (GMM) and ensemble-based techniques) in enhancing the performance of machine learning pipelines. The models were trained on a real industrial dataset compiled from power network analytics and harmonic-injection-based loading conditions. Model validation and equipment-level testing were performed using a large-scale harmonic measurement dataset collected over a five-year period. The reliability of the approach was confirmed by comparing predicted state transitions with actual fault occurrences, demonstrating its practical applicability and suitability for integration into predictive maintenance frameworks. The analysis demonstrates that correction functions introduce deterministic transformations in the data or error space, whereas improvement algorithms apply adaptive optimisation to fine-tune model parameters or decision boundaries. The combined use of these approaches significantly reduces overfitting, improves predictive accuracy and lowers false alarm rates. This work introduces the concept of an Organically Adaptive Predictive (OAP) ML model. The proposed model presents organic adaptivity, continuously adjusting its predictive behaviour in response to dynamic variations in network loading and harmonic spectrum composition. The introduced terminology characterises the organically emergent nature of the adaptive learning mechanism.