In recent decades, digital transformations have significantly impacted knowledge production and dissemination, especially in Islamic sciences, leading to issues such as digital religious content proliferation, text fragmentation, and misuse of fatwas. This study aims to present an integrated framework entitled "Intelligent Knowledge Engineering for Islamic Texts" to address these challenges through a maqasid-based and contextual approach. The research tackles cognitive challenges in digital religious content by establishing theoretical foundations for Islamic knowledge engineering that leverage generative artificial intelligence while respecting maqasid principles. It also proposes an applied model linking texts to their objectives and contexts through semantic processing techniques. The study adopts an analytical-descriptive methodology, reviewing Islamic and linguistic sources and analyzing digital religious data using AI tools. Findings indicate that employing AI in Islamic research enhances fatwa accuracy, organizes digital religious content, and supports maqasid-oriented ijtihad in contemporary issues. The study also provides a set of guidelines for the effective and safe application of this intelligent knowledge engineering framework. Keywords: Islamic text, knowledge engineering, maqasid, generative artificial intelligence, digital religious content

Abstract The scarcity of annotated Single-Photon Emission Computed Tomography (SPECT) images limits the performance of data-driven medical imaging models. Although SPECT is routinely employed in the clinical detection of bone metastases, its low spatial resolution and high noise hinder the generation of diverse, anatomically accurate training samples. To address these challenges, we propose Frequency-Aware SPECT-Conditioned Diffusion, a conditional diffusion framework for high-fidelity SPECT image generation. By incorporating frequency-aware modules, our model enhances anatomical accuracy while promoting lesion diversity, enabling the synthesis of structurally precise and pathologically varied images. Frequency-Aware SPECT-Conditioned Diffusion develops two novel components into the reverse denoising process: a High-Frequency Transform Attention (HFTA) module that restores crucial anatomical detail by enhancing high-frequency information at the bottleneck, and a Frequency Domain Enhancement Module (FDEM) that enriches global-local feature representation through frequency decomposition along the skip connections. Extensive experiments on real clinical SPECT datasets demonstrate that Frequency-Aware SPECT-Conditioned Diffusion achieves superior quantitative results compared with existing generative models, achieving an MSE of 13.42, PSNR of 32.99 dB, and SSIM of 0.8264. Moreover, the synthesised images produced by our method improve downstream tasks, boosting classification F1-score from 0.6850 to 0.7533 and segmentation Dice from 0.4143 to 0.4469. These results demonstrate the dual effectiveness of Frequency-Aware SPECT-Conditioned Diffusion in both high-quality image synthesis and enhanced task-specific model generalization in SPECT-based applications.

Multivariate flow dynamics-conditioned diffusion for automated structural optimization of semi-filled micro gas chromatography columns.

Exploration of the assessment of clinical decision-making capabilities in Clinical Oncology based on generative large language models.

The extraordinary progress of strategies coupling ab initio calculations and machine learning (ML) has opened the door for both fast and accurate chemical/physical property predictions and for the virtual design of materials. However, these techniques are very often used as a "black box" with the sole objective of obtaining high accuracy with scarce or no special attention on how ML models obtain their predictions. This can be improved by leveraging explainability of ML models, which, at the same time, would increase the chance of ML to offer new insights into the chemistry and physics of materials. Hence, the next generation of ML models in these realms must guarantee explainability by embedding explainable artificial intelligence (XAI) tools into their pipelines. Specifically, ML-assisted materials discovery and design can take great advantage of the use of XAI. Enabling explanations would increase the impact of these approaches by providing not only a set of candidates, but also insights into what makes a given material better than others. With this in mind, using the example of heterogeneous catalysts for hydrogen production and energy generation, here we propose a novel strategy for materials design based on counterfactual explanations. We were able to find materials featuring properties close to the design targets that were later validated with density functional theory calculations. Explanations were devised by comparing original samples, counterfactuals, and discovered candidates. Such explanations allowed us to unveil subtle relationships between the most relevant features, other, in principle, less important features, and the target property. Since this approach can be applied to different applications, this work provides an alternative to already available designing strategies, such as high-throughput screening or generative models, but that, for the first time, incorporates explainability as its main driving mechanism.

Generative models have been widely adopted in the biomedical domain, especially in image generation applications. Latent diffusion models achieve state-of-the-art results in generating brain MRIs. However, due to latent compression, generated images from these models are overly smooth, lacking fine anatomical structures and scan acquisition noise that are typically seen in real images. We propose image-to-image diffusion models that are designed to enhance the realism and details of generated brain images by introducing sharp edges, fine textures, subtle anatomical features, and imaging noise. This work formulates the realism enhancing and detail adding process as an image-to-image diffusion model, which refines the quality of LDM-generated images. We employ commonly used metrics like FID and LPIPS for image realism assessment. Furthermore, we introduce new metrics to quantify the improved realism of images generated by RealDeal in terms of image noise distribution, sharpness, and texture.

G-NeuroDAVIS: A generative model for data visualization through a generalized embedding.

This review critically examines work at the intersection of machine learning (ML) and metal-organic frameworks (MOFs). The modular nature of MOFs enables immense design flexibility and applicability to a wide range of applications. However, the combinatorially large design space also stresses the resource-intensive nature of traditional high-throughput screening approaches. Due to the increasing availability of data in the form of experimental and hypothetical MOF structures and their properties, ML methods have emerged as a promising solution to accelerate MOF discovery, yet successful application of these methods will require strategies that maximize data and resource efficiency. This work surveys approaches to reduce data and resource burdens for MOF property prediction and design through feature engineering, model architecture choices, transfer learning, active learning, and generative models. We also discuss challenges related to data quality and scalability, as well as future opportunities for ML-empowered methods that, up to this point, have primarily focused on MOF adsorption properties. By focusing on efficiency at every stage (from data generation to model inference), we identify future pathways for making ML-aided MOF design more robust and accessible to both theorists and experimentalists alike.

Multi-modal recommender system using text-to-image generative models and adaptive learning

Exploring diverse solutions in network dismantling with generative model

The bone marrow microenvironment(BME) maintains bone homeostasis through multi-cellular cooperation and signal crosstalk, itsdysregulation drivespathological boneloss. In recent years, Materiobiology, a scientific discipline studying how biomaterial properties affect biological functions, has opened new avenues for the precise regulation of this complex microenvironment. Biomaterials enable sophisticated regulation of the BME through biomimetic design and functionalization strategies. They not only activate osteoblast signaling pathways to promote bone formation but also inhibit osteoclast differentiation and bone resorption functions. Additionally, they integrate nerve and vascular regeneration processes with immunomodulatory mechanisms to optimize stem cell behavior and improve the tissue repair microenvironment. This review comprehensively summarizes advances in biomaterial-mediated BME regulation, emphasizing interdisciplinary integration and intelligent material development to overcome the limitations of conventional therapies. The innovation of intelligent materials lies in their ability to mimic biological systems. Recent research has leveraged generative design models to engineer new thiol-containing antimicrobial peptides. These approaches achieve spatiotemporal coordination of cellular interactions and functional reconstruction during bone regeneration. Future efforts need to address challenges in material stability, personalized adaptation, and clinical translation, promoting cross-scale therapeutic innovation from molecular intervention to tissue regeneration, providing revolutionary solutions for bone metabolic diseases and complex defect repair.

Controllable data augmentation and application of multiaxial fatigue experiments by fatigue conditional generative adversative network model

Performance and improvement strategies for adapting generative large language models for electronic health record applications: A systematic review.

Informed consent (IC) ensures patient understanding on proposed medical procedures, including endobronchial ultrasound (EBUS). Artificial intelligence (AI) presents as a potential tool to improve this process. This study explores the potential of AI to improve traditional IC documents and if AI-generated video consents are a feasible alternative. An AI-generated IC (AI-IC) was created using a generative AI model. In phase I, participants evaluated both AI-IC and traditional IC (H-IC) unidentified texts through a 5-point Likert scale questionnaire and selected their preferred. In phase II, patients answered a questionnaire evaluating the AI-generated IC in text (AI-IC) or video (AIV-IC) format. In phase I, (n=75, 44% health care professionals), AI-IC received higher scores for language clarity (P=0.013), benefits explanation (P<0.001), and addressing complications (P<0.001), but had lower scores for detailing the procedure (P<0.001). Most participants (86.7%) preferred the AI-IC for mentioning alternative procedures. In phase II, patients expressed high satisfaction with both the AI-IC (n=8) and AIV-IC (n=12). AIV-IC was globally accepted for replacing verbal IC. AI-generated materials improve accessibility in the IC for EBUS. While human supervision remains essential, future studies could strengthen the integration of AI-assisted and video-based consent tools in clinical practice.

A generative physics-informed machine learning model for soil microplastic accumulation dynamics.

In this methods article, we provide a flexible but easy-to-use implementation of direct coupling analysis (DCA) based on Boltzmann machine learning, together with a tutorial on how to use it. The package adabmDCA 2.0 is available in different programming languages (C , Julia, Python) usable on different architectures (single-core and multicore CPU, GPU) using a common front-end interface. In addition to several learning protocols for dense and sparse generative DCA models, it allows to directly address common downstream tasks like residue-residue contact prediction, mutational-effect prediction, scoring of sequence libraries, and generation of artificial sequences for sequence design. It is readily applicable to protein and RNA sequence data.

This study explores low-dimensional active subspace of pre-trained VAE-based generative molecular design models to perform uncertainty-guided fine-tuning of the VAE parameters to enhance downstream design task performance of the pre-trained models.

Abstract The limited coverage of radar sites has given rise to a demand for transforming the extensive coverage of weather satellite observations into high-resolution and accurate synthetic radar reflectivity imagery. In this study, we introduce a new method that utilizes generative diffusion models to address this challenge. Starting from pure noise, our diffusion model takes infrared images from the Himawari geostationary weather satellite and lightning observations from a ground-based network as inputs to control the generation process. The model’s iterative diffusion and denoising process helps capture the intrinsic uncertainty of satellite-to-radar transformation by generating probabilistic results, whereas nongenerative methods can only produce deterministic outputs. Our new technique improves the granularity and spatial accuracy of synthetic radar reflectivity imagery compared to previously published nongenerative U-Net models. In our experiments, the new technique enhances the emulation of severe weather by capturing finer visual structures in areas with strong radar echoes. Results show that images generated by our model outperform traditional U-Net models on key metrics such as the fractions skill score (FSS) across multiple thresholds, with the average FSS increasing from 0.40 to 0.50, and also produce a much improved statistical distribution of reflectivity, especially at the low and high ends of the distribution. Significance Statement Radar observations are essential for severe weather monitoring and nowcasting. However, the geographical limitations of radar coverage, particularly in Australia’s remote regions, present obstacles to providing radar-based warnings for high-impact weather events. This study introduces an innovative machine learning approach to transform widely accessible satellite infrared images and lightning observations into synthetic radar reflectivity images to overcome this limitation. Compared to previous methods, our approach improves the spatial resolution and accuracy of synthetic radar imagery, extending the applicability of severe weather nowcasting to regions beyond the reach of traditional radar systems. This advancement may improve severe weather monitoring and nowcasting in areas not covered by operational radar networks.

Generative deep learning models for cloud removal in satellite imagery: A comparative review of GANs and diffusion methods

Redefining medical visual question answering using conditional generative diffusion models