搜索结果：Gaussian

Controlled rotation of single biological cells is significant for cellular biology and engineering. Here we present a light-driven and non-contact strategy that enables arbitrary-axis rotation of both spherical and anisotropic cells with real-time switching between distinct rotation modes (major-axis and minor-axis). The platform employs a Bovine Serum Albumin (BSA)-coated gold nano-island (AuNIs) plasmonic film to generate strong interfacial thermo-osmotic flow under laser illumination, while Polyethylene Glycol (PEG)-induced depletion forces confine cells near the interface. For spherical particles, arbitrary-axis rotation is achieved using a single Gaussian beam, where spatial asymmetry in the thermo-osmotic flow determines the rotation axis. For anisotropic cells, different rotation modes are enabled by optical pattern reconfiguration. A Gaussian beam induces major-axis rotation, while a half-ring beam generates a combined optical and thermo-osmotic torque distribution that supports sustained minor-axis rotation. The rotation mode is reversibly switched solely through optical reconfiguration without mechanical intervention. This unified platform establishes geometry-independent, optically programmable rotational control, opening new opportunities for high-speed multi-angle cellular imaging and dynamic studies of cell-cell interactions.

Post-traumatic stress disorder (PTSD) is maintained by dysfunctional trauma-related appraisals. Cognitive Bias Modification for Appraisals (CBM-APP) aims to train more functional trauma-related appraisals and has been shown to reduce PTSD symptoms. However, little is known about how this training affects the interrelations among symptoms and cognitive appraisals. In this secondary analysis of a randomized controlled trial involving 77 adult patients diagnosed with PTSD (CBM-APP: n = 37; control training: n = 40), we applied repeated cross-sectional network analysis to examine changes in the structure and centrality of associations among PTSD symptom clusters (re-experiencing, avoidance, negative cognition and mood, hyperarousal, assessed with the PTSD Checklist for DSM-5) and trauma-related cognitive measures, all assessed at both pre- and post-training. To capture multiple levels of cognitive processing, we included responses during a scenario task (reflective, idiosyncratic, spontaneous appraisals) and the Implicit Association Test (automatic self-associations). Four cross-sectional Gaussian Graphical Models were estimated for the training and control group and both timepoints (pre-/post-training × CBM-APP vs. control group). While overall network connectivity did not differ significantly across networks, descriptive patterns indicated that Alterations in Cognition and Mood emerged as the most central node in both groups at post-training assessment. Further, in the CBM-APP group, the centrality of implicit trauma-related associations decreased pre- to post-training, suggesting potential weaker associations of automatic negative self-associations with symptom activation. Given the small sample and moderate network stability, findings are preliminary but suggest that CBM-APP may influence the relational structure of PTSD symptoms and cognitions.

Recently, the zero-shot image captioning (zero-shot IC) method based on pre-trained visual language models (VLMs) and large language models (LLMs) has made significant progress. However, how to adapt it to the zero-shot video captioning (zero-shot VC) scenario (without video-text paired supervision) has not been well explored. Inspired by various recent test-time strategies (sacrificing additional test time to improve performance), we try to introduce a new paradigm of Test-time Reinforcement Polishing in zero-shot VC scenario.We take temporal dependency modeling as the starting point and propose a novel framework for Refinable Zero-shot VC, called RefZVC. RefZVC can greatly cover the long-term context of the video and continuously polish and refine the generated captions in a reward-feedback manner. We first design an Adaptive Frame Skipping module (AdaSkip) to skip redundant frames and select diverse keyframe sequences. Subsequently, we propose a Multi-granularity Reinforcement Polishing (MRP) mechanism, which iteratively polishes captions by leveraging Gaussian Kernel Cache (GKC) to capture temporal dynamics, store and reuse relevant historical context. In addition, MRP calculates rewards for generated captions at both the sentence-level and entity-level to achieve test-time polishing. With the MRP mechanism, RefZVC achieves superior zero-shot generalization performance, outperforming previous zero-shot VC methods on benchmarks such as MSVD, MSR-VTT, and VATEX.

Three novel D‑π1‑π2‑A organic dyes with polyheterocyclic π2‑bridges and cyano‑modified acceptors were designed to amplify the polarity contrast between the terminal groups. Density functional theory (DFT) calculations were used to characterize the energy levels, absorption spectra, and intramolecular charge transfer (ICT) characteristics, with a focus on the regulatory effects of different electron‑withdrawing substituents on electron injection, dye regeneration, and charge recombination at the dye@TiO2 interface. Simulation results indicate that SGT-149-3 exhibits the best photovoltaic performance, which can be attributed to two key structural optimizations: the replacement of the π2-bridge with dithienothiophene and the rational modification of the acceptor with cyano group. The narrow energy gap leads to a significant redshift of 112.21 nm in the absorption spectrum compared to the reference dye SGT-149. The enhanced ICT effect improves the separation efficiency of photogenerated carriers, while stronger covalent interactions with TiO2 substrates and prolonged fluorescence lifetime suppress charge recombination. The short-circuit current density and photoelectric conversion efficiency of SGT-149-3 reach 28.53 mA/cm2 and 25.831%. Such exceptional photovoltaic parameters validate the latent application in photovoltaic devices. This work reveals the structure-property relationship between π-bridge engineering and acceptor modification, providing guidance for the rational design of high-efficiency dye-sensitized solar cells (DSSCs) sensitizers. The ground-state geometries of all dye molecules were optimized at the B3LYP/6-311 g(d,p) level of DFT. The excited-state properties were calculated using the CAM-B3LYP/6-311 g(d,p) level of TD-DFT. For the dye@TiO2 and dye@I systems, the 6-311 g(d,p) basis set was used for light atoms (C, H, O, N, S), while the LANL2TZ(f) pseudopotential basis set was applied to Ti and I atoms. The geometries of these complexes were optimized at the B3LYP level. All the simulations were conducted in THF solution using the SMD solvent model. This study employed the Gaussian 16 software package as the core computational tool. Multiwfn_3.8 was used to analyze the π-electron delocalization and charge transfer characteristics.

Elucidating reaction mechanisms requires efficient generation of transition states (TSs) and products. Existing diffusion and sequence-based models accelerate parts of this process over traditional string-based methods, but typically still require manual enumeration of either TSs or products, and stochastic diffusion dynamics can be inefficient and hard to control. We introduce MolGEN, a conditional flow-matching framework that uses deterministic optimal transport to map Gaussian priors to chemical distributions. For TS generation, MolGEN improves TS geometry and barrier-height prediction over diffusion models while enabling sub-second sampling. For reaction product generation, it achieves competitive top-k accuracy while preserving mass and electron balance. Using the same backbone for TS and product sampling, MolGEN enables template-free generative exploration of reaction networks without the repeated quantum-chemistry searches required by prior methods. For the γ-ketohydroperoxide decomposition network, it produces more valid TSs than string-based methods using only 12 quantum-chemistry evaluations instead of 1156, and identifies a lower-barrier pathway.

In semi-arid regions, daily relative humidity exhibits strong nonlinear and regime-dependent variability, which limits the robustness of conventional predictive approaches. This study investigates the influence of regularization on predictive stability and generalization within two supervised learning frameworks: Support Vector Regression (SVR) and Radial Basis Function Neural Networks (RBF-NN), using long-term meteorological observations from Fez, Morocco (1985-2022). Seven atmospheric predictors representing thermodynamic, radiative, and moisture-related processes were used as inputs. For SVR, hyperparameters were optimized using Bayesian optimization with cross-validation. While RBF-NN parameters and regularization were determined using grid-search and hold-out validation. Both models were evaluated using an identical data partitioning scheme (70% training, 15% validation, 15% testing) to ensure a consistent comparison. The results indicate that regularization effects differ across the two frameworks. In SVR, the Gaussian RBF kernel achieves the highest predictive performance (R = 0.9890; MSE = 0.0016), reflecting strong representation of relative humidity dynamics under global regularization through parameter C. In contrast, the RBF-NN model exhibits higher sensitivity to regularization parameter selection due to its localized learning structure, with the optimal configuration (λ = 0.01; architecture [7-16-1]) with (R = 0.9603; MSE = 0.0141), while deviation from this range leads to overfitting or underfitting. These results demonstrate that SVR provides more stable predictive performance than RBF-NN under identical climatic conditions. The study contributes to the understanding of how model structure and regularization influence relative humidity prediction under semi-arid climatic conditions.

Quantitative gait analysis is an important tool in clinical assessment and biomechanical research. Vision-based motion capture systems used in dedicated gait laboratories are considered the reference standard, providing highly accurate kinematic measurements, but they require specialized infrastructure and controlled environments. As a result, wearable inertial sensor systems have become a practical alternative for clinical and everyday gait assessment. This study presents a comparative evaluation of a smartphone-based gait analysis method and a clinically established wearable inertial system.
Gait recordings representing healthy, mildly asymmetric, and impaired walking patterns were analysed using both approaches. The comparison focused on clinically relevant parameters, including cadence, swing-phase duration symmetry, and vertical acceleration symmetry. The smartphone-based method employs median-based statistics and robust outlier rejection to enhance reliability under non-ideal measurement conditions. 
For healthy and mildly asymmetric gait, both systems produced comparable parameter estimates. In gait patterns with pronounced imbalance, the smartphone-based method detected larger deviations, indicating enhanced sensitivity to asymmetry and pathological alterations, highlighting the benefits of robust statistical modelling when gait data deviate from Gaussian distributions.
These findings demonstrate a strong level of agreement between the proposed smartphone-based method and a clinically established wearable inertial system while supporting the practical applicability and robustness of the approach for quantitative gait assessment.
The method provides gait parameters comparable to those obtained with a wearable reference system while maintaining stable performance across different walking conditions. Its low instrumentation requirements and applicability outside laboratory settings 
make it suitable for rehabilitation follow-up, fall risk assessment, and repeated gait evaluation over time.

Lactococcus lactis SH6 was engineered for heterologous heparosan production, and its growth medium was optimized using a combination of experimental design and machine learning (ML). One-factor-at-a-time shake flask experiments revealed glucose (10 g/L) and yeast extract (17.5 g/L) as the best substrates, producing 50 mg/L heparosan. Plackett-Burman analysis and steepest ascent optimization revealed significant factors, and a central composite design (CCD) optimized nutrient concentrations, predicting 85 mg/L heparosan (validated at 81 mg/L). ML-Gaussian process regression was applied after CCD optimization to fine-tune and cross-check the optimal medium (glucose 8.94 g/L, yeast extract 22.89 g/L, ascorbate 0.38 g/L, β-glycerophosphate 28.2 g/L), producing 85.28 mg/L heparosan (predicted 88.8 mg/L) at the flask scale. Earlier nisin induction (2 h) at the bioreactor scale increased heparosan titers to 119.7 mg/L, and linear glucose feeding (1.5 g/L.h) extended the production phase to 133 mg/L. Medium optimization resulted in nearly doubling heparosan yield compared to the unoptimized medium, setting a new standard for L. lactis. This work offers a design-of-experiments-ML solution as a viable approach to designing high-yielding, animal-product-free heparosan production methods in a Generally Regarded as Safe (GRAS) microbe.

Achromatic structured focusing in flat optics is important for compact depth-resolved imaging and three-dimensional localization, yet simultaneously achieving broadband operation and a depth-encoding focal field remains challenging. Here, we demonstrate an achromatic double-helix metalens operating across the visible range of 600-650 nm. The target phase profile combines an exact hyperbolic lens phase with a double-helix phase synthesized from Laguerre-Gaussian mode bases, whose coefficients are optimized by particle swarm optimization. Broadband implementation is achieved through a meta-atom library designed by jointly considering phase delay, group delay, and group delay dispersion. The fabricated device exhibits strong structural fidelity and experimentally reproduces the rotating two-lobe focal field at 625 nm as well as across the target wavelength range. This work extends achromatic metalenses beyond conventional focusing to broadband structured focusing and provides a compact platform for depth-resolved imaging.

Wavelike motion mediated by chemotactic signaling occurs in various biological phenomena including neutrophil swarms, wound healing, and amoeba aggregates. However, the macroscopic transition from independent to collective cellular behavior remains unclear, including how to quantify the response of individual cells to a developing chemotactic wave. Recent advances in molecular imaging allow concurrent observation of cyclic adenosine monophosphate (cAMP) concentrations and cell movement at individual resolution. Employing particle image velocimetry (PIV), a scheme to extract Eulerian velocity vector fields in fluids, we derived velocity fields at different Gaussian blurring levels and found that while original fluorescent images reflect cell movement, blurred versions highlight cAMP wave propagation. We identified the phase of cAMP signal wave dynamics and analyzed the interplay between single-cell motility and cAMP wave development. The extracted velocity fields at single-cell resolution show an almost antipodal relationship to those of the cAMP wave characterized by the blurred image, with an angle close to 180° during the rise of the wave when the spiral wave is well developed. Furthermore, single-cell dynamics collectively move toward the crest of the coming cAMP wave but rest (with randomized directionality) in the troughs between waves, akin to "surfing of the collectives".

Galectin-3 is a β-galactoside-binding protein involved in multiple biological processes, including cell proliferation, apoptosis, and inflammatory response. The pivotal role played by galectin-3 in diverse cellular processes makes it a therapeutic target of choice against cancers. However, the dynamic interplay of galectin-3 with diverse carbohydrate ligands is yet to be fully understood in detail. Elucidating molecular recognition mechanism(s) of ligand binding in a system such as galectin-3 is crucial for biological function and has promising implications in drug discovery. However, capturing these rare-event mechanisms using atomistic molecular dynamics simulations is a computationally expensive task. Our work aims to uncover the nuances of one such crucial mechanism using Gaussian accelerated molecular dynamics (GaMD) with galectin-3 as a prototypical system. The current work employs GaMD to explore the intricate binding mechanisms of galectin-3 with glycans of varying chain lengths. Our results show that pentasaccharides and ortho-fluoro derivatives as longer-chain glycans have interacted with galectin-3 strongly and stably, recapitulating the cocrystal pose. These ligands interact not only with the CRD of galectin-3 but also remain bound in the native pocket over longer time scales, in contrast to shorter-chain ligands such as beta-lactose and TF-antigen. However, due to inherent conformational flexibility, shorter-chain ligands were observed to sample allosteric pockets prior to reaching the native binding mode. This suggests that the length of the galectin-3 ligand plays a crucial role in determining its mode of interaction. This study advances the understanding of galectin-3 recognition and provides important considerations for designing effective galectin-3-targeting therapeutics.

Cloud service providers face the challenge of determining optimal server replenishment policies that minimize inventory costs while ensuring the expected demand satisfaction rate. This article addresses a long-term, single-echelon inventory optimization problem tailored to the unique characteristics of cloud, incorporating varying given demand satisfaction rates for different cloud products. Traditional replenishment methods, based on statistical approaches or conventional reinforcement learning (RL) algorithms, are ill-suited due to the uncertainties in user demand retention and the diversity in demand satisfaction requirements. To overcome these limitations, we propose an enhanced replenishment policy under given demand satisfaction rates using RL, combined with our policy imitation. Specifically, we introduce a data-driven safety stock (DSS) model as an expert policy, utilizing Kullback-Leibler (KL) divergence and advantage values for effective action imitation. Based on the results of DSS, we design an efficient reward function and propose an approach for hyperparameter estimation. Together, they enable our method to learn better policies that align with the given demand satisfaction rate. We further design an end-to-end network that integrates demand retention prediction with RL, incorporating a learnable Gaussian kernel to model the lead time effects and local features, and a mean self-attention module for capturing global features. To accelerate training, we also introduce a parallel algorithm. Extensive experiments on real-world data from multiple internet data centers (IDCs), across different time scales and given demand satisfaction rates, demonstrate the superior performance of our method. Additionally, ablation studies and parameter sensitivity analyses verify the efficacy of our policy imitation method and network architecture.

Thermal infrared imaging is pivotal for all-weather 3D perception, yet analyzing thermal information remains a formidable challenge due to the complexity of heat conduction. Unlike visible light, heat conduction acts as a natural low-pass filter that suppresses high-frequency textural details, causing severe geometric ambiguities and "ghosting" artifacts in standard 3D reconstruction pipelines. To accurately model the inherently diffusive thermal field for high-fidelity reconstruction, we propose Frequency-Gated Graph Splatting (ThermalGate-GS), a frame work that explicitly decouples the scene into diffusive thermal distributions (low-frequency) and sharp structural boundaries (high-frequency). Within this framework, we introduce a novel Frequency-Gated Anisotropic Diffusion mechanism. Specifically, the frequency-gating module utilizes extracted high-frequency structural cues to determine spatially-adaptive gating weights. Subsequently, these weights drive an anisotropic diffusion process that dynamically regulates thermal feature propagation, pro moting smoothness on object surfaces while suppressing cross-boundary bleeding. Finally, these spectrally refined features are employed to regress 3D Gaussian attributes, substantially alleviating the ambiguity in thermal reconstruction. Extensive experiments demonstrate that ThermalGate-GS achieves state-of-the-art performance, with a notable 7.94 dB PSNR improvement on the ThermoScenes benchmark over prior physics-inspired baselines.

Analyzing the complex and diverse soundscapes of ecosystems such as coral reefs remains a challenge for understanding environmental dynamics and processes. While machine learning techniques can significantly improve detection and classification capabilities, applications of traditional supervised learning to underwater acoustics are limited by the size and class-coverage of labeled datasets. Unsupervised machine learning offers the potential to detect and classify sounds without the guidance of human labels, including signals that were unknown to the human analyst. However, the majority of previously developed unsupervised approaches characterize reef soundscapes from correlative metrics without identifying specific sounds, and the few that detect individual signals have been trained on limited data (<10 days), which constrains the potential to generalize across datasets and geographical localities. Here, a convolutional autoencoder was built and trained on year-long acoustic datasets from four Hawaiian coral reefs, and latent embeddings were clustered using Gaussian mixture modeling. A total of 29 classes were automatically generated, and a manual review of samples in each class determined that nine of the classes corresponded to distinct biological and anthropogenic sounds. The classes were identified to be two call types from the damselfish, Dascyllus albisella, parrotfish feeding sounds, holocentrid calls, an unidentified fish sound, three humpback whale song units, and ship noise. The classifier was found to be robust against an independently-collected test dataset with D. albisella calls (AUC&#x2009;=&#x2009;0.9) with no extra training on the labels. Diel, lunar, and seasonal trends were observed for all nine classes, including previously-unidentified responses of the holocentrid and unknown fish groups to lunar illumination. This work demonstrates the capability of unsupervised algorithms to cluster acoustic signals into identifiable biological and anthropogenic categories in order to examine and characterize ecological trends.

The transition from late adolescence to emerging adulthood is a critical period of psychological and neurobiological change. Gaming disorder (GD), depression, and anxiety often co-occur, yet little is known about how their comorbidity structures reorganize across developmental stages. A two-wave longitudinal study was conducted with 663 Chinese adolescents assessed during late adolescence (T&#x2080;, mean age&#x2009;=&#x2009;17.52) and 620 participants reassessed three years later in emerging adulthood (T&#x2081;, mean age&#x2009;=&#x2009;20.53). Validated self-report scales measured symptoms of GD, depression, and anxiety. Symptom networks were estimated separately at each wave using regularized Gaussian graphical models. Expected influence (EI) and bridge expected influence (BEI) were computed to assess central and bridging symptoms, network stability was evaluated using bootstrap procedures, and Network Comparison Tests were conducted to compare global strength and network structure between waves. From T&#x2080; to T&#x2081;, network sparsity decreased from 0.80 to 0.75, density increased from 0.20 to 0.25, and global strength increased from 19.55 to 22.50. At T&#x2080;, the highest EI nodes were GAD1 (nervousness, 1.23) and GD6 (persistent gaming, 1.11), and the highest BEI nodes were GAD1 (0.50) and PHQ6 (worthlessness, 0.41). At T&#x2081;, the highest EI nodes were GAD5 (restlessness, 1.22) and GAD3 (excessive worry, 1.19), and the highest BEI nodes were GAD5 (0.64) and PHQ9 (suicidal thoughts, 0.62). Network Comparison Tests indicated that global EI increased by 1.28 (p&#x2009;=&#x2009;0.02). Centrality invariance testing revealed significant EI increases for GAD3 (0.58, p&#x2009;<&#x2009;0.01) and GAD5 (0.57, p&#x2009;<&#x2009;0.01). For BEI, GAD7 (catastrophizing, 0.42, p&#x2009;=&#x2009;0.02) and GAD3 (0.40, p&#x2009;<&#x2009;0.01) showed significant increases. The longitudinal network analysis showed that the comorbidity of GD, depression, and anxiety reorganized from late adolescence to emerging adulthood. Cognitive-affective symptoms became central and bridging, whereas gaming-related emotional and behavioral links weakened, reflecting a possible developmental shift from reactive, behavior-based coping in adolescence to internally sustained, cognitively driven symptom networks in emerging adulthood. Interventions should focus on strengthening adaptive coping and targeting core cognitive-affective symptoms to reduce long-term psychopathological risk. Not applicable.

Predicting cognitive decline from brain MRI is a central question in neuroscience. Hippocampal volume (HV) is a key cognitive biomarker, and normative models can be augmented with multimodal information. Here we augment normative models with genetic information and show improvements in cognitive decline prediction across multiple experimental setups. We improve normative models for HV by integrating multi-threshold polygenic scores (PGS) with demographic and imaging data using Gaussian Process Regression (GPR). Models were trained on 23,997 participants from UK Biobank (UKBB) and validated on 3,000 out-of-sample participants from the Alzheimer's Disease Neuroimaging Initiative (ADNI) and the European Prevention of Alzheimer's Disease (EPAD) cohorts. Our genetically-informed models significantly strengthened associations across six experimental designs and 13 key neurocognitive measures, including Mini-Mental State Examination (MMSE), Clinical Dementia Rating (CDR), and Alzheimer's Disease Assessment Scale (ADAS), while enhancing prediction of future cognitive decline. Together, these findings underscore the promise of integrating multi-threshold PGS with neuroimaging-based predictive models to improve prognostication and early intervention strategies for neurodegenerative diseases.

Plant diseases significantly affect agricultural productivity and global food security, while accurate disease identification remains challenging because of uncertain and overlapping visual symptoms in leaf images. Existing deep learning approaches often require large annotated datasets and suffer from limited interpretability in practical agricultural environments. This study presents a multi-parameter improved fuzzy soft set-based framework for plant disease classification using tomato leaf images from the PlantVillage dataset. The objective is to develop an interpretable and reliable classification model capable of handling uncertainty in plant disease patterns through feature-driven fuzzy similarity analysis. The methodology integrates image preprocessing, color and texture feature extraction, variance-based feature weighting, prototype generation using K-means clustering, and fuzzy similarity computation using Mahalanobis distance and Gaussian membership functions. RGB, HSV, and Gray-Level Co-occurrence Matrix (GLCM) features are extracted from standardized leaf images and evaluated within an improved fuzzy soft classification framework. Performance comparison is carried out using machine learning models including Support Vector Machine (SVM), Random Forest (RF), Linear Discriminant Analysis (LDA), and Naive Bayes (NB) implemented in Python using Scikit-learn libraries. Experimental simulation results demonstrate that the proposed framework achieves competitive classification performance while preserving interpretability and robustness under uncertain feature distributions. Performance evaluation is conducted through accuracy analysis, ROC-AUC curves, confusion matrices, ablation studies, and Wilcoxon Signed-Rank statistical testing. The proposed Improved Fuzzy Soft model achieved an accuracy of 88.57% which is less than LDA (94.92%), Random Forest (97.78%) and SVM (97.94%) classifiers. However, in the cross data set validation, the proposed Improved Fuzzy Soft model achieved an accuracy of 67.35% which is greater than LDA (51.02%), Random Forest (51.02%) and SVM (55.10%) classifiers. Statistical validation using the Wilcoxon Signed-Rank Test produced a p-value of [Formula: see text], confirming that the performance difference between the Improved Fuzzy Soft framework and the Random Forest classifier is statistically significant under the current experimental setting.

Epoxy-based nanocomposites are promising solid insulation materials for high-voltage applications because of their high dielectric strength, mechanical robustness, and processability. However, identifying the optimal nanofiller loading that maximizes dielectric breakdown strength (BDS) remains challenging because conventional trial-and-error approaches are costly, time-consuming, and difficult to generalize across material systems. This study proposes a cross-material physics-informed machine learning framework integrating four structurally distinct nanofillers: Zn/Al-LDH, Mg/Al-LDH, &#x3b3;-Al2O3, and &#x3b1;-Al2O3, where the alumina nanoparticles were synthesized from recycled aluminum beverage cans as a sustainable material source. For each system, 15 breakdown measurements per concentration were statistically validated using Weibull analysis. Among all systems, &#x3b1;-Al2O3 exhibited the highest BDS of 46.8&#xa0;kV/mm at 5 wt%, corresponding to approximately 56% improvement over neat epoxy. Dataset augmentation was performed using PCHIP interpolation combined with Gaussian noise injection and validated through Leave-One-Out Reconstruction analysis, which showed interpolation errors below 7.76% for internal concentration points. Composite relative permittivity at intermediate concentrations was estimated using the Maxwell-Garnett model and incorporated as a physically constrained input feature. Five regression algorithms were trained and benchmarked on material-specific datasets, achieving R2 values up to 0.959 with experimental validation errors below 5.2%. A cross-material model was further developed by replacing categorical material identity with intrinsic filler permittivity as a physics-based descriptor, enabling a transferable across multiple investigated nanofiller systems using a common descriptor. The cross-material model achieved R2 = 0.919 with prediction errors below 6%. The proposed framework provides a scalable and experimentally validated route for optimizing epoxy insulation systems for GIS/GIL spacer applications.

Phenylsilane is an organosilicon molecule containing a silyl group attached to a benzene ring and is widely used as a hydrogen source in organic transformations such as the Mukaiyama hydration reaction. Understanding how external electric fields influence its structural and spectroscopic properties is important for clarifying field-induced molecular behavior. In this work, the effects of external electric fields on the C-C and Si-C bond lengths, total energy, infrared (IR) spectrum, Raman spectrum, UV-vis absorption spectrum, and bond dissociation energy of phenylsilane are investigated. The results show that increasing the electric field strength induces redshifts in several vibrational absorption peaks, including a shift of up to 250&#xa0;cm - 1 for the Si-H vibrational mode, and leads to the appearance of new spectral features. At the maximum applied field strength, the Si-C bond length decreased from 1.882 to 1.842 &#xc5;, while the HOMO-LUMO gap decreased from 5.41 to 3.87 eV. In addition, several Raman peaks become more distinct with increasing electric field intensity. The bond length, total energy, UV-vis absorption characteristics, and the bond dissociation energy also exhibit systematic changes under different electric field strengths. All calculations were performed using density functional theory (DFT) at the B3LYP/6-31+G(d) level of theory. Geometry optimizations and vibrational frequency analyses were carried out under different external electric field strengths. Infrared and Raman spectra as well as UV-vis absorption spectra were obtained based on the optimized molecular structures. All quantum chemical calculations were performed using the Gaussian 16W software package.