Accurate fault detection in high-dimensional industrial environments remains a major challenge due to the inherent complexity, noise, and redundancy in sensor data. This paper introduces CLAIRE, i.e., a hybrid end-to-end learning framework that integrates unsupervised deep representation learning with supervised classification for intelligent quality control in smart manufacturing systems. It employs an optimized deep autoencoder to transform raw input into a compact latent space, effectively capturing the intrinsic data structure while suppressing irrelevant or noisy features. The learned representations are then fed into a downstream classifier to perform binary fault prediction. Experimental results on a high-dimensional dataset demonstrate that CLAIRE significantly outperforms conventional classifiers trained directly on raw features. Moreover, the framework incorporates a post hoc phase, using a game-theory-based interpretability technique, to analyze the latent space and identify the most informative input features contributing to fault predictions. The proposed framework highlights the potential of integrating explainable AI with feature-aware regularization for robust fault detection. The modular and interpretable nature of the proposed framework makes it highly adaptable, offering promising applications in other domains characterized by complex, high-dimensional data, such as healthcare, finance, and environmental monitoring.
Hazem Amamou, Stéphane Gagnon, Alan Davoust
et al.
Retrieval-Augmented Generation (RAG) was introduced to enhance the capabilities of Large Language Models (LLMs) beyond their encoded prior knowledge. This is achieved by providing LLMs with an external source of knowledge, which helps reduce factual hallucinations and enables access to new information not available during pretraining. However, inconsistent retrieved information can negatively affect LLM responses. The Retrieval-Augmented Generation Benchmark (RGB) was introduced to evaluate the robustness of RAG systems under such conditions. In this work, we use the RGB corpus to evaluate LLMs in four scenarios: noise robustness, information integration, negative rejection, and counterfactual robustness. We perform a comparative analysis between the RGB RAG baseline and GraphRAG, a knowledge graph based retrieval system. We test three GraphRAG customizations to improve robustness. Results show improvements over the RGB baseline and provide insights for designing more reliable RAG systems for real world scenarios.
In the evolutionary computation community, it is widely believed that stagnation impedes convergence in evolutionary algorithms, and that convergence inherently indicates optimality. However, this perspective is misleading. In this study, it is the first to highlight that the stagnation of an individual can actually facilitate the convergence of the entire population, and convergence does not necessarily imply optimality, not even local optimality. Convergence alone is insufficient to ensure the effectiveness of evolutionary algorithms. Several counterexamples are provided to illustrate this argument.
Sleep is essential for maintaining human health and quality of life. Analyzing physiological signals during sleep is critical in assessing sleep quality and diagnosing sleep disorders. However, manual diagnoses by clinicians are time-intensive and subjective. Despite advances in deep learning that have enhanced automation, these approaches remain heavily dependent on large-scale labeled datasets. This study introduces SynthSleepNet, a multimodal hybrid self-supervised learning framework designed for analyzing polysomnography (PSG) data. SynthSleepNet effectively integrates masked prediction and contrastive learning to leverage complementary features across multiple modalities, including electroencephalogram (EEG), electrooculography (EOG), electromyography (EMG), and electrocardiogram (ECG). This approach enables the model to learn highly expressive representations of PSG data. Furthermore, a temporal context module based on Mamba was developed to efficiently capture contextual information across signals. SynthSleepNet achieved superior performance compared to state-of-the-art methods across three downstream tasks: sleep-stage classification, apnea detection, and hypopnea detection, with accuracies of 89.89%, 99.75%, and 89.60%, respectively. The model demonstrated robust performance in a semi-supervised learning environment with limited labels, achieving accuracies of 87.98%, 99.37%, and 77.52% in the same tasks. These results underscore the potential of the model as a foundational tool for the comprehensive analysis of PSG data. SynthSleepNet demonstrates comprehensively superior performance across multiple downstream tasks compared to other methodologies, making it expected to set a new standard for sleep disorder monitoring and diagnostic systems.
A. V. Lobanov, E. D. Makeeva, D. O. Meshkov
et al.
Further development of manned space exploration requires appropriate scientific and appropriate modern methodological approaches assessing the health risks of participants in missions to an orbital station with high inclination of the orbit and interplanetary missions. The "Environmental Scanning" as well as available Internet sources (PubMed and EMBASE; 198 relevant publications) and personal messages, followed by expert discussion within the framework of an interdisciplinary working group were used to assess the contemporary approaches of manned spaceflight risk definition, classification, assessment and management. The results indicated that the term "risk" itself needs to be clarified, and mathematical models based on modern approaches other than differential calculus and providing expert support for management decisions should be developed or adapted to available evidence-based real-world data obtained in experiments using laboratory animals, ground-based simulation studies with the participation of volunteers, as well as during pre- and post-flight examinations of astronauts. It is advisable for this purpose to use interdisciplinary and interdepartmental working groups, including experts in the field of aviation, space and marine medicine, public health and health organization and control sciences as well as contemporary mathematical methods of analysis and statistics.
Synthesizing high-quality photorealistic images with textual descriptions as a condition is very challenging. Generative Adversarial Networks (GANs), the classical model for this task, frequently suffer from low consistency between image and text descriptions and insufficient richness in synthesized images. Recently, conditional affine transformations (CAT), such as conditional batch normalization and instance normalization, have been applied to different layers of GAN to control content synthesis in images. CAT is a multi-layer perceptron that independently predicts data based on batch statistics between neighboring layers, with global textual information unavailable to other layers. To address this issue, we first model CAT and a recurrent neural network (RAT) to ensure that different layers can access global information. We then introduce shuffle attention between RAT to mitigate the characteristic of information forgetting in recurrent neural networks. Moreover, both our generator and discriminator utilize the powerful pre-trained model, Clip, which has been extensively employed for establishing associations between text and images through the learning of multimodal representations in latent space. The discriminator utilizes CLIP's ability to comprehend complex scenes to accurately assess the quality of the generated images. Extensive experiments have been conducted on the CUB, Oxford, and CelebA-tiny datasets to demonstrate the superiority of the proposed model over current state-of-the-art models. The code is https://github.com/OxygenLu/RATLIP.
Heterogeneity arising from label distribution skew and data scarcity can cause inaccuracy and unfairness in intelligent communication applications that heavily rely on distributed computing. To deal with it, this paper proposes a novel personalized federated learning algorithm, named Federated Contrastive Shareable Representations (FedCoSR), to facilitate knowledge sharing among clients while maintaining data privacy. Specifically, the parameters of local models' shallow layers and typical local representations are both considered as shareable information for the server and are aggregated globally. To address performance degradation caused by label distribution skew among clients, contrastive learning is adopted between local and global representations to enrich local knowledge. Additionally, to ensure fairness for clients with scarce data, FedCoSR introduces adaptive local aggregation to coordinate the global model involvement in each client. Our simulations demonstrate FedCoSR's effectiveness in mitigating label heterogeneity by achieving accuracy and fairness improvements over existing methods on datasets with varying degrees of label heterogeneity.
Given the Amharic language has a lot of perplexing terminology since it features duplicate homophone letters, fidel's ሀ, ሐ, and ኀ (three of which are pronounced as HA), ሠ and ሰ (both pronounced as SE), አ and ዐ (both pronounced as AE), and ጸ and ፀ (both pronounced as TSE). The WSD (Word Sense Disambiguation) model, which tackles the issue of lexical ambiguity in the context of the Amharic language, is developed using a deep learning technique. Due to the unavailability of the Amharic wordnet, a total of 1756 examples of paired Amharic ambiguous homophonic words were collected. These words were ድህነት(dhnet) and ድኅነት(dhnet), ምሁር(m'hur) and ምሑር(m'hur), በአል(be'al) and በዢል(be'al), አቢይ (abiy) and ዐቢይ(abiy), with a total of 1756 examples. Following word preprocessing, word2vec, fasttext, Term Frequency-Inverse Document Frequency (TFIDF), and bag of words (BoW) were used to vectorize the text. The vectorized text was divided into train and test data. The train data was then analysed using Naive Bayes (NB), K-nearest neighbour (KNN), logistic regression (LG), decision trees (DT), random forests (RF), and random oversampling technique. Bidirectional Gate Recurrent Unit (BiGRU) and Bidirectional Long Short-Term Memory (BiLSTM) improved to 99.99 % accuracy even with limited datasets.
In this paper we study the transmission of higher-order modes, including optical vortices (OVs), through a bus fiber evanescently coupled with a vertical array of ring resonators (VAR), which form a vertically stacked multi-ring resonator. It is shown that the OV transmission curves have a characteristic structure that we explain by the manifestation of the band structure of an infinite stack of coupled ring resonators. We demonstrate a fundamental possibility of using VARs as elements of delay lines for fiber-optic communications using orbital angular momentum. It is shown that the VAR is capable of serving as a delay line element for even and odd Laguerre–Gauss modes.
D. Kh. Nurligareev, I. A. Nedospasov, K. Yu. Kharitonova
Objectives. The purpose of the article is to carry out a theoretical and experimental study of the angular reflection spectrum of linearly polarized electromagnetic waves from a multilayer periodic mirror on a transparent substrate to exact analytical expressions for reflection and transmission coefficients generalizing the cases of incidence of plane transverse electric (TE) and transverse magnetic (TM) modes on limited periodically structured media with a stepped refractive index profile.Methods. The theoretical analysis of the reflection problem is based on the search for exact analytical solutions in the form of Floquet–Bloch waves presented in the form of inhomogeneous waves in the domain of periodically structured media. On the basis of the possible existence of a single Floquet–Bloch wave in a limited onedimensional photonic crystal, it is proposed to search for exact solutions of the wave equation in the form of a linear combination of inhomogeneous waves propagating in different directions. By using the canonical forms of the considered periodic structures, it is possible to carry out the simple transition from the case of TE polarization to TM type in dispersion relations and expressions for the angular reflection spectrum.Results. Cases of reflection of linearly polarized radiation are considered for the following cases: a flat boundary of two dielectrics, a thin plane-parallel plate, and a multilayer dielectric mirror. Exact analytical expressions for the reflection and transmission coefficients generalizing the cases of incidence of TE and TM polarizations waves on a limited one-dimensional photonic crystal are obtained. The transmission coefficients of a plane TE wave from a multilayer dielectric mirror sputtered on thin glass were experimentally measured.Conclusions. A quantitative and qualitative agreement of experimental measurements of the transmission coefficient of a plane wave incident from a half-space on a confined photonic crystal with theoretical calculations is obtained. The obtained expressions for the transmission coefficient of a confined one-dimensional photonic crystal, which are shown to be determined by the interference of Floquet–Bloch waves presented in the form of inhomogeneous waves, can be reduced to a form analogous to the expression for the value of the transmission coefficient of a traditional Fabry–Pérot interferometer. In the case of TM polarization, when the Brewster condition is fulfilled at the interlayer boundaries, the Floquet–Bloch wave has the form of homogeneous plane waves in the layers of a photonic crystal.
Predicting molecular properties is crucial in drug synthesis and screening, but traditional molecular dynamics methods are time-consuming and costly. Recently, deep learning methods, particularly Graph Neural Networks (GNNs), have significantly improved efficiency by capturing molecular structures’ invariance under translation, rotation, and permutation. However, current GNN methods require complex data processing, increasing algorithmic complexity. This high complexity leads to several challenges, including increased computation time, higher computational resource demands, increased memory consumption. This paper introduces InvarNet, a GNN-based model trained with a composite loss function that bypasses intricate data processing while maintaining molecular property invariance. By pre-storing atomic feature attributes, InvarNet avoids repeated feature extraction during forward propagation. Experiments on three public datasets (Electronic Materials, QM9, and MD17) demonstrate that InvarNet achieves superior prediction accuracy, excellent stability, and convergence speed. It reaches state-of-the-art performance on the Electronic Materials dataset and outperforms existing models on the R2 and alpha properties of the QM9 dataset. On the MD17 dataset, InvarNet excels in energy prediction of benzene without atomic force. Additionally, InvarNet accelerates training time per epoch by 2.24 times compared to SphereNet on the QM9 dataset, simplifying data processing while maintaining acceptable accuracy.
The global energy sector is undergoing a transformative shift driven by the urgent need to transition to clean and digital energy solutions. Start-ups are at the forefront of this transformation, providing innovative technologies to tackle critical challenges in energy production, distribution, and consumption. This study explores the influence of market entry registration procedures and regulatory frameworks on the growth and success of start-ups in the clean and digital energy sectors, highlighting their role in driving global energy transitions. This research aims to understand how procedural complexity, regulatory quality, and associated costs impact entrepreneurial activity in clean and digital energy industries. Employing a comprehensive methodology based on panel data analysis, the study examines multi-country datasets from authoritative sources, such as the International Energy Agency and the World Bank, spanning two decades. Fixed and random effects models are utilized to reveal nuanced relationships between regulatory conditions and start-up proliferation. By identifying key barriers and enablers, the research provides actionable insights into fostering an environment conducive to innovation and entrepreneurship. The findings underscore several critical aspects of market entry dynamics. Streamlined registration procedures emerge as a pivotal factor in promoting start-up growth, reducing the administrative burden and enabling quicker market access. Conversely, high capital requirements and extended registration timelines act as significant deterrents, limiting the ability of new ventures to scale effectively. Regulatory quality is shown to play a crucial role in fostering an innovation-friendly environment, with higher-quality frameworks positively correlated with entrepreneurial success. Additionally, the study reveals the synergistic effects of entrepreneurial ecosystems, where broader business activity enhances opportunities for start-up development. These results highlight the dual nature of regulatory frameworks, which can either facilitate or hinder start-up activity. While effective regulations provide necessary oversight and ensure market stability, overly burdensome procedures can stifle innovation and deter market entry. Policymakers are urged to strike a balance by simplifying procedural requirements, reducing capital thresholds, and maintaining robust regulatory oversight to foster a thriving ecosystem for clean and digital energy start-ups. The study’s contribution is particularly timely, given the accelerating pace of global energy transitions and the increasing focus on achieving sustainability goals. The actionable insights offered here can guide policymakers and stakeholders in creating regulatory environments that catalyze innovation and entrepreneurship. By enabling start-ups to navigate market entry challenges effectively, governments can harness their potential to drive technological advancements and contribute meaningfully to a sustainable energy future. Future research directions include cross-country comparative analyses to identify best practices and longitudinal studies to examine the long-term impacts of regulatory optimization on entrepreneurial ecosystems. The findings of this study provide a foundation for ongoing efforts to align regulatory practices with the goals of a decarbonized and digitally integrated energy landscape, ensuring a resilient and inclusive energy transition.
Owing to advancements in deep learning technology, Vision Transformers (ViTs) have demonstrated impressive performance in various computer vision tasks. Nonetheless, ViTs still face some challenges, such as high computational complexity and the absence of desirable inductive biases. To alleviate these issues, {the potential advantages of combining eagle vision with ViTs are explored. We summarize a Bi-Fovea Visual Interaction (BFVI) structure inspired by the unique physiological and visual characteristics of eagle eyes. A novel Bi-Fovea Self-Attention (BFSA) mechanism and Bi-Fovea Feedforward Network (BFFN) are proposed based on this structural design approach, which can be used to mimic the hierarchical and parallel information processing scheme of the biological visual cortex, enabling networks to learn feature representations of targets in a coarse-to-fine manner. Furthermore, a Bionic Eagle Vision (BEV) block is designed as the basic building unit based on the BFSA mechanism and BFFN. By stacking BEV blocks, a unified and efficient family of pyramid backbone networks called Eagle Vision Transformers (EViTs) is developed. Experimental results show that EViTs exhibit highly competitive performance in various computer vision tasks, such as image classification, object detection and semantic segmentation. Compared with other approaches, EViTs have significant advantages, especially in terms of performance and computational efficiency. Code is available at https://github.com/nkusyl/EViT
In recent times, significant advancements have been made in delving into the optimization landscape of policy gradient methods for achieving optimal control in linear time-invariant (LTI) systems. Compared with state-feedback control, output-feedback control is more prevalent since the underlying state of the system may not be fully observed in many practical settings. This paper analyzes the optimization landscape inherent to policy gradient methods when applied to static output feedback (SOF) control in discrete-time LTI systems subject to quadratic cost. We begin by establishing crucial properties of the SOF cost, encompassing coercivity, L-smoothness, and M-Lipschitz continuous Hessian. Despite the absence of convexity, we leverage these properties to derive novel findings regarding convergence (and nearly dimension-free rate) to stationary points for three policy gradient methods, including the vanilla policy gradient method, the natural policy gradient method, and the Gauss-Newton method. Moreover, we provide proof that the vanilla policy gradient method exhibits linear convergence towards local minima when initialized near such minima. The paper concludes by presenting numerical examples that validate our theoretical findings. These results not only characterize the performance of gradient descent for optimizing the SOF problem but also provide insights into the effectiveness of general policy gradient methods within the realm of reinforcement learning.