In-situ monitoring methods and deep learning models are increasingly being used for the quality assessment of parts fabricated using laser powder bed fusion to overcome the limitations of poor process repeatability. However, the massive data collection required for part-quality monitoring results in high transmission loads and storage costs. To address this problem, this study utilized the compressed sensing theory to acquire compressed photodiode signals. These signals were then used to train and test convolutional neural networks (CNN) to identify the lack-of-fusion, normal, and keyhole modes. At a compressive-sampling rate of 25%, the classification accuracy decreased from 93.1% (raw signals) to 79.3%. However, increasing the compression rate from 25% to 90% did not significantly decrease the classification accuracy. The linear mapping of the raw signal via a Gaussian measurement matrix causes coordinate information folding, thereby impairing the representation of latent features. Therefore, Gaussian process modeling was adopted for the features extracted using a pretrained CNN to mitigate the temporal information collapse and allow the compressed signals to achieve an accuracy comparable to that of the raw data. Furthermore, the sparsity and rank complexity of the melt-pool radiation signals were evaluated using sparse representation and principal component analysis.

Memory-efficient neural network training via gradient compression through continuous basis tracking

Training neural networks for the <mml:math xmlns:mml="http://www.w3.org/1998/Math/MathML" altimg="si38.svg" display="inline" id="d1e10529"> <mml:mrow> <mml:mi>G</mml:mi> <mml:mi>I</mml:mi> <mml:mo>/</mml:mo> <mml:mi>G</mml:mi> <mml:mo>/</mml:mo> <mml:mi>K</mml:mi> </mml:mrow> </mml:math> queue

Generation of synthetic labeled datasets for anomaly detection in heritage architecture

Hepatitis B virus (HBV) infection can lead to hepatocellular carcinoma, and HBV integration into the host genome is regularly observed in the liver of chronic HBV carriers and is speculated to trigger carcinogenesis. To detect HBV integration, PCR-based methods are sensitive but may not be effective depending on the HBV integration site. High-throughput sequencing reveals not only HBV integration but also the site. However, it is still expensive for clinical applications. In situ hybridization is efficient and allows detection of DNA and RNA in single cells and has been applied to study HBV infection and dynamics. The technique, however, has not been applied to detect HBV integration. The task is challenging because of the small copy number of integrated HBV genomes, which results in a small signal-to-noise ratio. Here, we developed a fluorescence in situ hybridization approach for examining HBV integration in each liver cell. The obtained images of cells were analyzed using a deep learning model. Using several hepatoma cell lines with and without integrated HBV DNA, we showed that our trained neural network achieved an over 90% accuracy for majority of positive and negative control cells. This is the first proof-of-concept study showing that fluorescence imaging and deep learning are useful in detecting HBV integration at the single-cell level.

Patch-wise learning is a common strategy for training neural networks on large-scale dense prediction problems, yet existing approaches assume uniform or fixed sampling distributions. This assumption is suboptimal when learning difficulty varies spatially and evolves with the model state during optimization. We reformulate patch-wise learning as a dynamic computation allocation problem and propose an adaptive patch sampling (APS) algorithm that learns where to sample by constructing model-state-dependent sampling distributions from voxel-wise uncertainty and prediction error. To learn what contextual information is encoded within sampled patches, we introduce a patch encoding (PE) block that infers implicit location information and modulates feature representations through context-dependent channel-wise attention, without relying on explicit spatial coordinates. Experiments on whole-body multi-organ PET-CT segmentation demonstrate faster convergence and consistent performance gains, with external validation on Synapse dataset confirming robustness. Mechanistic analyses of learning dynamics further characterize sampling behavior induced by APS and representation modulation driven by the PE block through attention analysis and causal channel pruning. Overall, this work contributes an efficient learning strategy for patch-wise training and provides insight into how dynamic sampling and contextual conditioning influence optimization in large-scale dense prediction tasks.

The structural dynamics of DNA underpin essential biological processes, yet conventional structural biology methods often obscure conformational heterogeneity through ensemble averaging. Atomic force microscopy (AFM) provides single-molecule topographical maps capable of capturing both local and global variation, but extracting quantitative insight from these images remains challenging. Here, we introduce an automated framework that reduces AFM data to spline representations of the DNA backbone and applies cyclic Procrustes analysis to quantify shape similarity across ensembles. Using purified topoisomers of 339bp DNA minicircles ranging from relaxed to highly negatively supercoiled, we resolved and measured the relative abundance of conformational states across the different topoisomers, capturing gradual transitions among open circles, compact conformations, and self-crossing structures that are invisible to techniques such as gel electrophoresis or cryoelectron microscopy (cryo-EM). We show that beyond quantification, Procrustes distances provide supervisory signals for neural network training, enabling feature extraction tuned to conformational geometry and supporting robust conformation classification of AFM images. Extending the same spline representation to molecular dynamics simulations allows experimental and computational ensembles to be directly compared, establishing a common shape-based framework for probing conformational variability. Together, these advances transform AFM from a descriptive imaging tool into a quantitative platform for mapping conformational continua, with broad applicability to DNA and other dynamic biomolecular systems.

The clinical feasibility and translation of many advanced quantitative MRI (qMRI) techniques are inhibited by their restriction to 'research mode', due to resource-intensive, offline parameter estimation. This work aimed to achieve 'clinical mode' qMRI, by real-time, inline parameter estimation with a trained neural network (NN) fully integrated into a vendor's image reconstruction environment, therefore facilitating and encouraging clinical adoption of advanced qMRI techniques. The Siemens Image Calculation Environment (ICE) pipeline was customized to deploy trained NNs for advanced diffusion MRI parameter estimation with Open Neural Network Exchange (ONNX) Runtime. Two fully-connected NNs were trained offline with data synthesized with the neurite orientation dispersion and density imaging (NODDI) model, using either conventionally estimated (NNMLE) or ground truth (NNGT) parameters as training labels. The strategy was demonstrated online in two healthy volunteers (one rescanned) and evaluated offline with synthetic data, testing two diffusion protocols. NNs were successfully integrated and deployed natively in ICE, performing inline, whole-brain, in vivo NODDI parameter estimation in < 10 s. The proposed workflow was reproducible across protocols, volunteers and rescans. DICOM parametric maps were exported from the scanner for further analyses. Comparisons between NNMLE and NNGT suggested NNMLE parameter estimates to be more consistent with conventional fitting, a finding supported by offline evaluations. Real-time, inline parameter estimation with the proposed generalizable framework resolves a key practical barrier to the potential clinical uptake of advanced qMRI methods, enabling their efficient integration into clinical workflows. Next steps include incorporation of pre-processing methods and evaluation in pathology.

The rapid proliferation of sophisticated cyber threats has exposed critical limitations in conventional security architectures that rely on isolated, reactive tools. This paper presents ZeroGuardian-XDR, an intelligent and lightweight Extended Detection and Response (XDR) framework engineered to deliver real-time network threat detection, automated vulnerability assessment, and proactive incident alerting through a unified platform. The proposed system employs a trained autoencoder neural network for behavioral anomaly detection, enabling the identification of zero-day and previously unknown threats without reliance on static signature databases. ZeroGuardian-XDR integrates nine live global threat intelligence feeds including AlienVault OTX, Abuse.ch, Feodo Tracker, URLhaus, Blocklist.de, ThreatFox, NVD CVEs, MITRE ATT&amp;CK, and EmergingThreats, collectively maintaining over 22,000 dynamic threat indicators automatically refreshed every six hours. The system maps all detections to the MITRE ATT&amp;CK framework with 87% technique coverage across 8 tactical phases and 691 monitored techniques. A professional SOC-style web dashboard, multi-channel alert delivery via Telegram and email, automated PDF report generation, and an Nmap-powered CVE vulnerability scanner complete the integrated architecture. Experimental evaluation using five simulated zero-day attack scenarios demonstrated 100% detection accuracy with minimal false positive rates. The framework is deployed on Ubuntu Server 24.04 and made publicly available through open-source distribution with Windows and Linux installer packages. ZeroGuardian-XDR represents a scalable, cost-effective, and academically reproducible cybersecurity solution for modern network protection

Data-driven neural networks (NNs) have gained significant attention across engineering disciplines, particularly in design optimization and experimental settings, where they are widely used to construct surrogate models for high-dimensional regression problems. Despite their power as global approximators, neural networks often struggle to accurately capture local features without relying on a large number of trainable parameters and training data points, resulting in increased training time. To address these limitations, in this paper we propose domain decomposition methods (DDMs), which divide the input feature space into multiple local subdomains, each modeled by a simpler NN, trained in parallel. Interface constraints are introduced in the local loss functions to enforce continuity between subdomains. They are enforced with two different approaches: by utilizing Lagrange multipliers or augmented Lagrange multiplier methods. Compared to unconstrained approximations, both methods significantly improve continuity across subdomain interfaces. For a 2D and a 3D problem, computational time and accuracy are investigated across varying numbers of subdomains to identify optimal partitioning strategies. The use of DDMs improves approximation accuracy in local regions with smaller number of parameters when compared to standard global NN training. In terms of convergence, the augmented Lagrange method outperforms the standard Lagrange formulation by converging faster due to lower convergence requirements, albeit with a slightly lower accuracy. Overall, these results highlight the augmented Lagrange method as a promising DDM approach for training efficient and scalable NN surrogate models.

Sub-visible particle analysis using flow imaging microscopy combined with deep learning has proven effective in identifying particle types, enabling the distinction of harmless components such as silicone oil from protein particles. However, the scarcity of available data and severe imbalance between particle types within datasets remain substantial hurdles when applying multi-class classifiers to such problems, often forcing researchers to rely on less effective methods. The aforementioned issue is particularly challenging for particle types that appear unintentionally and in lower numbers, such as silicone oil and air bubbles, as opposed to protein particles, where obtaining large numbers of images through controlled settings is comparatively straightforward. In this work, we develop a state-of-the-art diffusion model to address data imbalance by generating high-fidelity images that can augment training datasets, enabling the effective training of multi-class deep neural networks. We validate this approach by demonstrating that the generated samples closely resemble real particle images in terms of visual quality and structure. To assess the effectiveness of using diffusion-generated images in training datasets, we conduct large-scale experiments on a validation dataset comprising 500,000 protein particle images and demonstrate that this approach improves classification performance with no observable downside. Finally, to promote open research and reproducibility, we publicly release both our diffusion models and the trained multi-class deep neural network classifiers, along with a straightforward interface for easy integration into future studies, at https://github.com/utkuozbulak/svp-generative-ai.

Well-designed and trained deep neural networks can solve inverse electromagnetic problems much faster than conventional solvers. However, they need a physics framework to ensure producing physically correct results. Since most physics-guided deep learning inverse solvers require substantial training with numerous epochs, each involving solving a forward problem, their accuracy and efficiency are largely defined by the utilized forward solver, which becomes a bottleneck for their practical training. Thus, a fast and accurate self-supervised deep learning forward solver is presented. The solver uses a physics-based framework that divides the domain into two regions: an interior region, which includes any scatterers, and an exterior region, which represents the background medium. A hybrid loss function, incorporating Maxwell's curl equation and integral equation with the well-defined scalar background's Green's function, is employed to guide the scattered field generated from the neural network, ensuring global and local accuracy. To verify the generality of the solver, it is trained on random objects and tested on realistic models, showing high global and local metrics accuracy. For example, more than 95% of testing cases using the proposed method achieve less than 0.15 root-mean-square error in the calculated scattered field and dielectric properties of the imaged domain compared to the ground truth. In contrast, two recent deep learning methods could only realize that level of accuracy for less than 50% of the tested cases. The reported method is 97% faster than conventional solvers, enabling the development of reliable deep-learning inverse solvers.

• Introduce energy-conserving neural network closure for turbulence. • Skew-symmetric term redistributes energy; negative definite term dissipates energy. • Outperforms standard machine learning models; delivers accurate long-time LES. • Neural network training procedure consistently yields accurate and stable LES. Machine learning-based closure models for large eddy simulation (LES) have shown promise in capturing complex turbulence dynamics but often suffer from instabilities and physical inconsistencies. In this work, we develop a novel skew-symmetric neural architecture as closure model that enforces stability while preserving key physical conservation laws. Our approach leverages a discretization that ensures mass, momentum, and energy conservation, along with a face-averaging filter to maintain mass conservation in coarse-grained velocity fields. We compare our model against several conventional data-driven closures (including unconstrained convolutional neural networks), and the physics-based Smagorinsky model. Performance is evaluated on decaying turbulence and Kolmogorov flow for multiple coarse-graining factors. In these test cases, we observe that unconstrained machine learning models suffer from numerical instabilities. In contrast, our skew-symmetric model remains stable across all tests, though at the cost of increased dissipation. Despite this trade-off, we demonstrate that our model still outperforms the Smagorinsky model in unseen scenarios. These findings highlight the potential of structure-preserving machine learning closures for reliable long-time LES.

Brain tumour segmentation is a key application of AI in neuroimaging. Recently, federated learning (FL) has emerged as a strategic and increasingly relevant paradigm in neural computing due to its ability to address key challenges in large-scale neural network training, such as data access, privacy, collaborative learning, and model robustness. However, its adoption is currently hindered by high communication costs and the heterogeneity of client data. In this study, we investigated an efficient FL framework for brain tumour segmentation based on communication-aware optimization. We evaluated FedWSOComp, which integrates sparsification, quantization, and entropy-based encoding, in combination with a 3D U-Net architecture under both homogeneous and heterogeneous data distributions. The multi-institutional FeTS 2024 dataset was employed and partitioned into independent and identically distributed (IID) and non-IID settings, with an independent test set of 67 patients. An overall of 18 configurations combined sparsification rates and quantization levels. Performance was measured using Dice Similarity Coefficient (DSC) and 95th percentile Hausdorff Distance (HD95). Experimental results demonstrated that aggressive compression caused severe degradation in segmentation quality, with HD95 exceeding 60 mm. In contrast, higher retention with finer quantization achieved the best balance between efficiency and accuracy, reaching a DSC and HD95 mm on the test set under non-IID conditions. The findings demonstrated that, when configured with moderate-to-fine quantization and high sparsification retention, FedWSOComp enabled accurate and communication-efficient federated brain tumour segmentation. This study provides quantitative evidence and practical guidance for the deployment of FL-based segmentation models in privacy-sensitive and bandwidth-constrained clinical settings. • Analyse the impact of FedWSOComp, an integrated strategy combining top-k sparsification, quantization, and entropy-based encoding. • The performance of 18 different configurations (including IID and non-IID) was evaluated systematically. • High retention (60%) with fine quantization (64 clusters) optimizes performance.

Community-based drug checking services are challenged in their ability to reliably detect low concentration adulterants that are increasingly present in the illicit drug supply. Spectral signatures from commonly used field instruments such as infrared spectrometers require careful analysis to identify characteristic features in a complex mixture. In this study, we train neural network models for the detection of bromazolam and para-fluorofentanyl, using infrared absorption data collected at a point-of-care drug checking service. The neural network models classified the two components with an F1-score of 0.88 for bromazolam and 0.89 for para-fluorofentanyl. In comparison, a random forest model optimized using the same data set had an F1-score of 0.66 for bromazolam and 0.76 for para-fluorofentanyl. This demonstrates that neutral networks are excellent candidates for such complex drug detection applications and outperform other machine learning-based approaches.

Weighted neural network and weighted least square estimators under censorship.

The development of digital humanities and intelligent audio analysis technology provides a new computational path for the emotional research of multi-ethnic folk songs. Taking the multi-ethnic Spring Festival ballads of Gansu Province as the object, this paper constructs a sentiment classification model by focusing on corpus collection, audio preprocessing, MFCC parameter extraction and LSTM neural network training. A total of 526 valid samples were sorted out, with a cumulative duration of 1149.2 minutes. The samples were labeled as four types of emotions: celebration, blessing, thinking and expressing, and narrative peace, and 39 dimensional MFCC temporal features were extracted as model input. Experimental results show that the model training loss decreases from 1.31 to 0.11, the training accuracy reaches 91.9%, and the validation accuracy reaches 85.4%. In the test set of 104 samples, the model correctly identified 90 samples, and the overall accuracy was 86.5%, the Precision, Recall and F1-score were 86.5%, 86.6% and 86.6%, respectively, which were better than SVM, CNN, RNN and GRU. The results show that the combination of MFCC and LSTM can effectively represent the emotional acoustic features in the Spring Festival songs, which provides technical support for the digital protection, emotional label construction and intelligent retrieval application of multi-ethnic Spring Festival songs in Gansu province.

This study aimed to develop and validate a physics-informed neural network (PINN) framework for data-efficient and physically consistent process optimization in the laser powder bed fusion (LPBF) of Inconel 718 (IN718) superalloy. Laser powder bed fusion (LPBF) is widely adopted for fabricating Inconel 718 (IN718) components in aerospace and energy applications; however, navigating its high-dimensional, nonlinear process parameter space remains a central challenge. High-fidelity finite element simulations are computationally prohibitive for extensive parameter sweeps, whereas purely data-driven machine learning (ML) models are limited by data scarcity and unphysical extrapolation behavior. This study presents a physics-informed neural network (PINN) framework that embeds the transient heat conduction equation and Goldak double-ellipsoidal heat source model directly into the neural network training loss, enforcing thermophysical consistency simultaneously with data fidelity. The model was trained on a curated, multi-source dataset of LPBF IN718 parameter combinations drawn from peer-reviewed experimental studies and validated finite element simulation outputs, spanning the laser power (70–400 W), scan speed (200–2000 mm/s), hatch spacing (50–140 µm), and layer thickness (20–50 µm). The PINN predicted the melt pool width, depth, peak temperature, and relative density with mean absolute percentage errors (MAPE) of 3.8%, 4.7%, 3.1%, and 1.9%, respectively, outperforming a baseline artificial neural network (ANN) with an identical architecture. The framework correctly identified the optimal volumetric energy density (VED) window of 55–105 J/mm3, yielding relative densities ≥99.5%, consistent with the published experimental thresholds for IN718. A data efficiency analysis demonstrated that the PINN with 25% training data achieves a performance equivalent to that of the fully trained ANN with 100% data, confirming an approximately four-fold data efficiency improvement attributable to physics-informed regularization, consistent with theoretical predictions. Sensitivity analysis via automatic differentiation confirmed that laser power and scan speed were the dominant parameters (~85% combined variance), which is in agreement with previous studies. This study provides a computationally efficient, interpretable, and physically consistent ML pathway for the accelerated process qualification of IN718 components for aerospace and energy applications.

The article examines current issue of legal qualifying use of copyright objects in artificial intelligence training. The author substantiates the need to amend the Civil Code of the Russian Federation by establishing a special case of free use of works for the purpose of training neural networks, including data collection. Based on the analysis of foreign experience and judicial practice, the author concludes that the use of works in the intellectual analysis of texts and data in digital form, including for the purpose of training neural networks, should be recognized as lawful provided that the form of the works is not perceived by human senses. It is proposed to extend this exception to any works in the public domain, including materials from the Internet and closed databases to which developers have obtained legal access. The paper substantiates the inexpediency of introducing a fee for the use of works in the process mentioned, as this may lead to a decrease in investment in technology development and complicate the process of training neural networks. At the same time, permissible and impermissible cases of use are clearly delimited: internal memorization of materials is not considered a violation, however, content generation with reproduction of significant parts of protected works is qualified as a violation of exclusive rights. It is substantiated the generation of works in the style of a particular author during neural network training based on his works may also constitute a violation of exclusive rights. Particular attention is paid to issues of liability for violations. The author proposes a differentiated approach according to which both the developer of the neural network and the user may be held liable, depending on the specific circumstances of the case. The study emphasizes the approach proposed will maintain a balance between protecting the rights of content creators and the need to develop AI technologies are important for solving global challenges in various spheres of public life.

Abstract In complex frictional systems, friction-induced vibration (FIV) and noise are ubiquitous and intricate issues. Achieving high-precision simulation of the vibration response is crucial for the diagnosis of system dynamic properties and vibration control. However, frictional surfaces with multiple contact points introduce nonsmoothness, resulting in unpredictable vibration responses and posing significant challenges for numerical methods to maintain accuracy over long-term analyses. This study proposes a new physics-informed neural network (PINN) method designed to enhance the adaptability between physical constraints and neural network training. The method introduces loss functions with state transition boundary modification (STBM) derived from the physical governing equations. Additionally, a data expansion and regression (DER) strategy for processing linear complementarity problem (LCP) is implemented in the optimizer, significantly improving simulation accuracy for complex stick–slip vibration processes in multicontact frictional systems. By combining these two innovations, the proposed method, referred to as BMDER-PINN, was validated through simulations of stick–slip vibration in a two-degree-of-freedom (2DoF) frictional system. Compared with conventional time-stepping methods, this approach ensures higher accuracy in longer simulations while also enabling large time steps, thereby offering a promising calculation method for improving nonsmooth dynamics simulations.