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.

A physics-embedded dual-learning imaging framework for electrical impedance tomography.

DLR-YOLO: Dynamic low-rank training for a lightweight power tower object detection network in multi-scenario remote sensing images

• 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.

• CLIP-driven rain perception enables adaptive routing for diverse rain patterns. • Mask-guided cross-attention enhances feature interaction between rainy and non-rainy regions. • Dynamic loss scheduling aligns with network optimization process, enhancing training effectiveness. • State-of-the-art performance achieved across multiple datasets, excelling in complex mixed scenarios. Existing deraining models process all rainy images within a single network. However, different rain patterns have significant variations, which makes it challenging for a single network to handle diverse types of raindrops and streaks. To address this limitation, we propose a novel CLIP-driven rain perception network (CLIP-RPN) that leverages CLIP to automatically perceive rain patterns by computing visual-language matching scores and adaptively routing to sub-networks to handle different rain patterns, such as varying raindrop densities, streak orientations, and rainfall intensity. CLIP-RPN establishes semantic-aware rain pattern recognition through CLIP’s cross-modal visual-language alignment capabilities, enabling automatic identification of precipitation characteristics across different rain scenarios. This rain pattern awareness drives an adaptive subnetwork routing mechanism where specialized processing branches are dynamically activated based on the detected rain type, significantly enhancing the model’s capacity to handle diverse rainfall conditions. Furthermore, within sub-networks of CLIP-RPN, we introduce a mask-guided cross-attention mechanism (MGCA) that predicts precise rain masks at multi-scale to facilitate contextual interactions between rainy regions and clean background areas by cross-attention. We also introduces a dynamic loss scheduling mechanism (DLS) to adaptively adjust the gradients for the optimization process of CLIP-RPN. Compared with the commonly used l 1 or l 2 loss, DLS is more compatible with the inherent dynamics of the network training process, thus achieving enhanced outcomes. Our method achieves state-of-the-art performance across multiple datasets, particularly excelling in complex mixed datasets.

Shadow-DETR: Alleviating matching conflicts through shadow queries.

Reinforcement-guided hyper-heuristic hyperparameter optimization for fair and explainable spiking neural network-based financial fraud detection

• Advanced estimation framework for on-board comfort from incomplete APC data. • Real-time performance of complex multi-line public transport networks. • Case study on Helsinki commuter train network with high estimation accuracy. • Reliable comfort estimates even at low APC coverage levels. • Guidance for optimal APC deployment and service quality monitoring. Comfort on-board public transport vehicles is a critical metric of user experience and service performance. The quantification of this metric requires knowledge of the number of passengers on-board every time a vehicle arrives at or departs from a stop or station. Automatic Passenger Counting (APC) systems allow obtaining such knowledge in real-time, but the information is often incomplete due to system malfunctions, or, more commonly, a lack of the relevant equipment in some vehicles. This study develops an advanced method for passenger estimation that fills gaps in incomplete APC datasets, with computational performance allowing real-time application, and calculates comfort levels on-board public transport vehicles in complex networks where stations are served by multiple lines. The proposed method is tested on a case study considering the Helsinki commuter train network, comprising 6 service lines and 20 stations. The results indicate that the proposed framework can achieve comfort level estimations with high precision across the different cases evaluated. Furthermore, the study provides insight into the key practical question of the number of vehicles that need to be equipped with APC devices in order to obtain sufficiently accurate on-board passenger comfort estimates, and it is shown that it is possible to obtain these estimates even when only a small subset of the runs of any single day are performed by equipped vehicles. Finally, the proposed estimation approach is a valuable tool for operators to obtain a better understanding of daily mobility patterns, evaluate their services through quantifying user experience, and enhance their operations.

Traditional side-channel analysis treats power and electromagnetic traces as temporal sequences, applying statistical or sequence-based machine learning methods without regard for the circuit topology responsible for generating observed leakage. This discards structural information intrinsic to digital circuits: gate connectivity, signal propagation topology, and the hierarchical organization of cryptographic modules. A framework is presented that applies Graph Neural Networks (GNNs) to side-channel vulnerability assessment by modeling circuits as attributed graphs in which nodes represent logic gates, edges represent wire connections, and power measurements are encoded as node features. A complete pipeline is developed spanning Verilog netlist parsing, graph construction, and Graph Convolutional Network (GCN) training with multi-head attention for multi-scale circuit analysis. Evaluation on ten AES-128 circuit implementations demonstrates an 86.4% attack success rate, compared with 68.1% for a CNN-LSTM baseline, with required power traces reduced from 988 to 790. Cross-architecture generalization reaches 63.5% accuracy on unseen circuit families, substantially above the 18.7% random baseline. Interpretable vulnerability heatmaps localize leakage sources at the gate level, enabling pre-silicon security assessment before fabrication.

Introduction In supervised classification tasks, models are trained to predict a label for each data point. In real-world datasets, these labels are often noisy due to annotation errors. While the impact of label noise on the performance of deep learning models has been widely studied, its effects on networks' hidden representations remain poorly understood. Method We address this gap by systematically comparing hidden representations using the Information Imbalance, a computationally efficient proxy of conditional mutual information. We analyze representations learned by networks of varying sizes, trained on datasets with controlled levels of label noise. Result Through this analysis, we observe that the information content of the hidden representations follows a double descent as a function of the number of network parameters, akin to the behavior of the test error. We show that in the underparameterized regime, representations learned with noisy labels are more informative than those with clean labels, while in the overparameterized regime they are equally informative. We also show that label noise decreases the information content between the penultimate and pre-softmax layers, mirroring the increase in the test error. Finally, representations learned from random labels perform worse than random features when the number of network parameters and training samples are scaled proportionally with a fixed ratio. Discussion Overall, our results suggest that representations of overparameterized networks are robust to label noise. The relationship between the information imbalance (between the penultimate and pre-spftmax layers) and the test error offers a new perspective on understanding generalization and highlights how training objectives shape internal representations. In addition, the poor performance of the representations learned from random labels—compared to random features—indicates that training in this setting goes beyond lazy learning.

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.

Abstract Dynamic (kinetic) models track time-varying metabolite concentrations, fluxes, and enzyme levels, quantifying responses to genetic and environmental perturbations. Yet building these models at scale is hindered by scarce enzyme kinetic parameters. Generative neural networks can rapidly parameterize near-genome-scale kinetic models, but their representations are hard to interpret and often require new training to move across species or physiological states. Here we introduce a latent-space exploration framework that repurposes a trained generative network to produce models with targeted dynamics in new regimes without additional training. We show in Escherichia coli that latent inputs tune aerobic response speed, identify rate-limiting enzymes, and retarget the generative network to anaerobic dynamics. We extend our approach to Saccharomyces cerevisiae , demonstrating robust control of metabolic dynamics across training stages and diverse latent inputs. Latent variables thus become practical control knobs for kinetic model behavior, accelerating cell-factory design and enabling personalized metabolic modeling.

To date, several Field-Programmed Gate Array (FPGA) implementable computational architectures have been proposed that can be used for neural network training in real-time by the backpropagation algorithm. However, they are intended for small neural networks or have a significant reduction in maximum clock frequency as network sizes increase. The novelty of this work lies in addressing the problems of ensuring a predictable maximum clock frequency and minimizing its degradation when scaling the computational architecture. The proposed architecture solves these problems at the level of computational organization. The architecture comprises an array of computational blocks which are based on FPGA digital signal processing blocks and perform most computations in parallel. The architecture also contains the shared block that sequentially processes the computation results received from the array blocks. The equations were derived showing that the latency of computations increases linearly with neural network sizes. After a computational block instance, the shared block and neural networks containing various numbers of computational blocks had been implemented on the FPGA, their timing characteristics were assessed. It has been determined that the data path delays of the buses connecting the shared block with the array blocks are the primary factors constraining the maximum clock frequencies of neural networks. When the number of the array blocks lies in the range 3–240, the maximum clock frequency is from 112 down to 77 MHz. Compared to the closest counterpart, the critical paths in the proposed architecture are shortened because some computations are transferred to the sequential mode; however, this transfer may increase the latency of calculating the local gradients of the hidden layers neurons. When the number of the array computational blocks grows from 3 to 128, the maximum clock frequency decreases by 27 % compared to 52 % for the closest counterpart. Growing the number of computational blocks in the proposed architecture from 128 to 240 reduces the maximum clock frequency by no more than 5 %. FPGA-based neural networks of the proposed architecture are suitable for object tracking and system identification, which are typical applications of neural networks trained in real-time mode.

The reliability of machines largely depends on the accuracy of predicting the stress–strain state of components in tribofatigue systems, especially under high operating loads. Traditional finite element analysis provides high accuracy but requires significant computational resources and offers limited flexibility for rapid parameter variation. In recent years, machine learning methods have been increasingly applied in engineering practice. Among them, neural networks are of particular interest, as they allow nonlinear relationships between loads and stresses to be captured while significantly reducing computation time compared to traditional models. This work proposes an approach for predicting maximum stresses in the “shaft–insert” system by combining three-dimensional finite element modeling with subsequent neural network training. A database was created containing the results of numerical experiments for different combinations of bending and contact loads. A fully connected neural network with three hidden layers and different activation functions was used for training. The quality of the model was assessed using standard metrics: Mean Squared Error, Mean Absolute Error (MAE), and the coefficient of determination R2. The trained neural network demonstrated high accuracy in predicting maximum stresses both in the shaft and in the insert. For the training set, the R2 value reached 0.99991, and for the test set it was 0.99984, confirming minimal deviations from finite element results. The MAE was less than 0.006, while the maximum relative error in the test set did not exceed 3.2 %. The developed neural network model demonstrated the ability to reproduce the results of finite element analysis for the “shaft–insert” system while providing a substantial reduction in computation time compared to traditional finite element simulations. The model was constructed for a limited range of loads; therefore, further research should focus on expanding the dataset and including additional materials, which will make it possible to evaluate the scalability of the approach and its robustness under more complex conditions.

ABSTRACT Climate‐induced flood risks pose a major challenge to sustainable development in Pakistan, particularly in the highly vulnerable province of Khyber Pakhtunkhwa. Recurrent flood events continue to undermine social, economic, and environmental systems, highlighting persistent difficulties in translating the Sendai Framework for Disaster Risk Reduction (SFDRR) into effective, context‐sensitive practice. This study examines the implementation barriers and enabling conditions for operationalizing the SFDRR in Khyber Pakhtunkhwa, situating provincial experiences within broader Global South sustainability challenges. Despite the global prominence of the SFDRR, limited empirical research has examined how its priorities are translated into operational disaster risk reduction practices at sub‐national governance levels in highly flood‐prone regions of the Global South. A qualitative research design was employed, drawing on 24 semi‐structured key informant interviews conducted between January and March 2025, guided by the four priorities of the SFDRR, which served as predefined analytical categories while allowing context‐specific sub‐themes to emerge inductively, and non‐participant field observations focused on flood‐prone areas, preparedness activities, and institutional coordination, triangulated with interview data. Participants were purposively selected based on their institutional roles, professional experience in disaster risk reduction, and direct involvement in flood management and disaster preparedness, drawing representatives from government institutions. Interviews were conducted until thematic saturation was reached. Data were analyzed using directed content analysis structured around the four SFDRR priorities, supported by methodological triangulation, independent coding, audit trails, and Lincoln and Guba's trustworthiness framework. Findings reveal that institutional fragmentation and weak inter‐agency coordination are the primary barriers, dynamically interacting with other constraints—including limited human and fiscal capacity, gaps in community engagement, and insufficient private‐sector participation—to create cascading vulnerabilities across flood‐affected communities. The study further identifies context‐specific enabling conditions and interdependent barriers unique to Khyber Pakhtunkhwa, offering a synthesized analytical framework that links institutional, technical, financial, and social constraints including gaps in Incident Command Structures, specialized training, and social network management for disaster communication and provides actionable guidance for prevention‐oriented, locally embedded disaster risk reduction. The study concludes that advancing flood resilience in Khyber Pakhtunkhwa requires a transition from reactive disaster response toward prevention‐oriented, institutionally embedded DRR aligned with the Sustainable Development Goals. (SDG 4.7, SDG 6.6, SDG 9.1, SDG 9.5, SDG 9.C, SDG 11.5, SDG 11.b, SDG 13.1, SDG 13.3, SDG 15.1, SDG 16.6, SDG 16.7, SDG 17.3, SDG 17.6, SDG 17.16, SDG 17.17). Policy implications emphasize integrated risk governance, sustained capacity building, resilient financing mechanisms, incentives for private sector engagement, coordinated response systems, and responsible digital communication to strengthen long‐term disaster resilience and sustainable development.

Abstract We propose a numerical method of searching for parameters with experimental constraints in generic flavor models by utilizing diffusion models, which are classified as a type of generative artificial intelligence (generative AI). As a specific example, we consider the $S_4^\prime$ modular flavor model and construct a neural network that reproduces quark masses, the CKM matrix, and the Jarlskog invariant by treating free parameters in the flavor model as generating targets. By generating new parameters with the trained network and local optimization, we find various phenomenologically interesting parameter regions. Additionally, we confirm that the spontaneous CP violation occurs in the $S_4^\prime$ model. The diffusion model enables an inverse problem approach, allowing the machine to provide a series of plausible model parameters from given experimental data.

Low-field magnetic resonance imaging (MRI) offers distinct advantages in terms of affordability, portability, and accessibility. However, its widespread adoption is limited by an inherently low signal-to-noise ratio and reduced spatial resolution. This study proposes an AI-assisted framework to enhance low-field MRI image quality and overcome these limitations. We propose a two-stage framework to generate high-quality low-field MRI images. In the first stage, realistic low-field images are synthesized from high-field acquisitions using a physics-informed forward model that incorporates spiral k-space trajectories and accounts for nonlinear magnetic field gradients, B0 inhomogeneity, k-space undersampling, and image reconstruction characteristics of low-field systems. In the second stage, a 3D U-Net enhanced with a multi-head attention in a Vision Transformer (ViT) module is trained on paired synthetic low-and high-field images to serve as a post processing following conventional image reconstruction. On the synthetic test set, our framework demonstrates strong performance, achieving a peak signal-to-noise ratio (PSNR) of 19.08 ± 2.85 dB for the baseline U-Net model and 21.00 ± 2.50 dB with the ViT block, demonstrating high reconstruction fidelity. The structural similarity index measure (SSIM) reaches 0.6456 ± 0.0779 (without ViT) and 0.6639 ± 0.0798 (with ViT), along with low normalized root mean squared error (NRMSE) values of 0.3866 ± 0.0952 and 0.3084 ± 0.0695, respectively. These results highlight significant improvements in both image quality and reconstruction robustness. The trained network, applied as a post-processing step after conventional reconstruction, consistently enhances the contrast-to-noise ratio (CNR) of the output images, supporting the qualitative observations of improved image contrast and clarity. The proposed framework addresses key limitations hindering the broader adoption of low-field MRI, including noise, artifacts, and resolution loss inherent to low-field acquisitions. By integrating deep learning with physics-based simulations, the approach achieves notable qualitative and quantitative enhancements in denoising, artifact removal, and overall image quality. These results highlight the framework's potential to improve the practical utility of low-field MRI substantially.

Over the years, Music Information Retrieval (MIR) research community has released various models pretrained on large amounts of music data. Transfer learning showcases the proven effectiveness of pretrained backend models for a broad spectrum of downstream tasks, including auto-tagging and genre classification. However, MIR papers generally do not explore the efficiency of pretrained models for Music Recommender Systems (MRS). In addition, the Recommender Systems community tends to favour traditional end-to-end neural network training. Our research addresses this gap and evaluates the performance of nine pretrained backend models (MusicFM, Music2Vec, MERT, EncodecMAE, Jukebox, MusiCNN, MULE, MuQ and MuQ-MuLan) in the context of MRS. We assess them using five recommendation approaches: K-Nearest Neighbours (KNN), Shallow Neural Network, Contrastive Multi-Modal projection, a Hybrid model, and BERT4Rec both for the hot and cold-start scenarios. Our findings suggest that pretrained audio representations exhibit significant performance disparity between traditional MIR tasks and both hot and cold music recommendations, indicating that valuable aspects of musical information captured by backend models may differ depending on the task. This study establishes a foundation for further exploration of pretrained audio representations to enhance music recommendation systems.

A Color Information Driven Collaborative Training of Dual Task Parallel Network for Visible and Thermal Infrared Image Fusion and Saliency Object Detection