Accurate Performance Prediction Research Articles

ABCC2 as a novel covariate influencing oxcarbazepine pharmacokinetics in pediatric epilepsy: Insights from population modeling.

This study set out to establish a population pharmacokinetic (PPK) model of oxcarbazepine (OXC) in pediatric patients with epilepsy, with the goal of optimizing individualized dosage strategies. We collected steady-state trough plasma concentration data of the OXC active metabolite 10-hydroxycarbazepine from epileptic children. The effects of physiological factors, concomitant medications, and genetic factors on pharmacokinetic parameters were quantitatively examined. Subsequently, a Monte Carlo simulation was carried out to predict steady-state trough concentration under various dosing regimens. A total of 320 plasma samples were obtained from 91 epileptic children. The results of covariate analysis revealed that body weight and ABCC2 rs2273697 had a significant impact on the clearance of the 10-hydroxycarbazepine. The final model demonstrated stable and accurate predictive performance, with simulation results indicating that a dosing regimen of 20-30 mg/kg per day is generally suitable for most pediatric patients to reach therapeutic targets. For children under 2 years, doses over 40 mg/kg per day might be needed. However, for those over 12 kg (2 years), 40-60 mg/kg per day could lead to excessive drug exposure. Patients carrying the ABCC2 variant allele necessitate a lower maintenance dose in comparison to those with the wild-type allele. This study marks the first instance to incorporate the genetic polymorphism of ABCC2 into a PPK model of OXC. The developed PPK model provides a fundamental basis for personalized dosing recommendations of OXC in epileptic children. SIGNIFICANCE STATEMENT: The relationship between 10-hydroxycarbazepine exposure and body weight, as well as the ABCC2 rs2273697 in epileptic children, can be accurately investigated using an age-stratified population pharmacokinetic model. Wild-type ABCC2 rs2273697 exhibited a 25% and 14% higher 10-hydroxycarbazepine apparent clearance compared to homozygous and heterozygous variation patients, respectively. Based on our model, an optimized dosing regimen for oxcarbazepine has been proposed aiming for various subgroups of patients.

Advanced biomaterial strategies for overcoming age-associated wound healing impairments.

Dermal wounds represent a substantial global healthcare burden, with significant economic impact and reduced quality of life for affected individuals. As skin ages, the wound healing capacity is significantly diminished through multiple pathways, including reduced cellular proliferation, altered inflammatory responses, impaired vascularization, and decreased extracellular matrix production. With worldwide demographics shifting toward an older population, effective wound management has become an increasingly critical healthcare challenge. Biomaterials have emerged as a powerful tool to address the specific challenges of wound healing by providing structural support and delivering therapeutic agents to facilitate tissue regeneration. These materials can even be engineered to match the specific mechanical properties of aged tissue while simultaneously releasing key age-tailored bioactive molecules, thereby addressing the complex healing deficits in aged skin. Recent advances in aged skin models have established them as crucial platforms for translational research, enabling more accurate prediction of biomaterial performance in elderly patients. Concurrently, composite biomaterials, which combine multiple functionalities in a single platform, have gained prominence as particularly promising clinical solutions. Though significant progress has been made, challenges persist in optimizing material properties and achieving reproducible clinical outcomes, demanding continued research focused specifically on age-related wound healing impairments.

Numerical Analysis of Lift Coefficients for NACA 4412 Airfoil Across Different Angles of Attack

The accurate aerodynamic performance prediction is critical for designing and optimising airfoils used in aerospace, automotive, and renewable energy applications. This study focuses on evaluating and comparing the lift coefficients of the NACA 4412 airfoil using three distinct methodologies: CFD, wind tunnel experimentation, and the Panel method. The primary objective is to assess each technique's accuracy and limitations in capturing the airfoil's aerodynamic characteristics. CFD simulations were conducted using ANSYS FLUENT, applying a steady-state, incompressible flow model with appropriate turbulence modelling to capture flow behaviour across a range of angles of attack. Experimental validation was performed in a controlled wind tunnel environment to generate benchmark data. The Panel method analysis was also executed using XFOIL, a commonly used inviscid flow solver known for its computational efficiency. The results demonstrate a strong agreement between CFD simulations and experimental data, particularly in predicting lift coefficients at moderate angles of attack. In contrast, XFOIL consistently overestimated lift values, especially at higher angles, due to its inability to accurately model flow separation and viscous effects. This discrepancy highlights the inherent limitations of potential flow methods when applied to complex flow regimes. The study systematically compares these approaches and emphasises the critical need for high-fidelity numerical or experimental validation when assessing airfoil performance. The findings advocate for a cautious application of simplified methods like the Panel method in preliminary design stages and reinforce the role of CFD as a reliable tool in aerodynamic analysis. This work contributes to the ongoing refinement of predictive tools for airfoil design, ensuring more accurate performance assessments in real-world applications.

Continuous integration of architectural performance models with parametric dependencies – the CIPM approach

The explicit consideration of the software architecture supports system evolution and efficient quality assurance. In particular, Architecture-based Performance Prediction (AbPP) assesses the performance for future scenarios (e.g., alternative workload, design, deployment) without expensive measurements for all such alternatives. However, accurate AbPP requires an up-to-date architectural Performance Model (aPM) that is parameterized over factors impacting the performance (e.g., input data characteristics). Especially in agile development, keeping such a parametric aPM consistent with software artifacts is challenging due to frequent evolutionary, adaptive, and usage-related changes. Existing approaches do not address the impact of all aforementioned changes. Moreover, the extraction of a complete aPM after each impacting change causes unnecessary monitoring overhead and may overwrite previous manual adjustments. In this article, we present the Continuous Integration of architectural Performance Model (CIPM) approach, which automatically updates a parametric aPM after each evolutionary, adaptive, or usage change. To reduce the monitoring overhead, CIPM only calibrates the affected performance parameters (e.g., resource demand) using adaptive monitoring. Moreover, a self-validation process in CIPM validates the accuracy, manages the monitoring to reduce overhead, and recalibrates inaccurate parts. Consequently, CIPM will automatically keep the aPM up-to-date throughout the development and operation, which enables AbPP for a proactive identification of upcoming performance problems and for evaluating alternatives at low costs. We evaluate the applicability of CIPM in terms of accuracy, monitoring overhead, and scalability using six cases (four Java-based open source applications and two industrial Lua-based sensor applications). Regarding accuracy, we observed that CIPM correctly keeps an aPM up-to-date and estimates performance parameters well so that it supports accurate performance predictions. Regarding the monitoring overhead in our experiments, CIPM’s adaptive instrumentation demonstrated a significant reduction in the number of required instrumentation probes, ranging from 12.6 % to 83.3 %, depending on the specific cases evaluated. Finally, we found out that CIPM’s execution time is reasonable and scales well with an increasing number of model elements and monitoring data.Graphical

The Prediction of Tensile Strength Performance of Fiber‐Reinforced Composites Based on a Knowledge Graph Constructed From Material Literature

ABSTRACTUnderstanding the relationship among material composition, processing methods, and mechanical properties plays an important role in the prediction of the tensile strength properties of polymer composites. Due to polymer composites' long and time‐consuming experimental cycles, obtaining data directly through experiments is challenging. At the same time, with the growing body of related literature, using literature mining and predictive methods to explore the relationships between material composition, processing methods, and mechanical properties has become a feasible alternative. Based on this, we propose an innovative approach that integrates data extraction, knowledge graph construction, and machine learning‐based performance prediction. A comprehensive dataset named ComMat for composites was constructed through systematic data sorting and filtering, encompassing nine entity categories and nine relationship categories representing key parameters such as materials, processing methods, testing, and performance. We proposed a literature data collection method that extracts structured triples from the literature using the joint extraction model PFPMHN, which was then used to build a domain‐specific knowledge graph. This knowledge graph enables the reverse inference of material performance and also supports feature selection through extensive reverse analysis, identifying key features related to material composition, processing parameters, and formulation. These features were used to train machine learning models, including XGBoost, achieving an R2 value of 0.96 for tensile strength prediction. Additionally, feature importance analysis, SHAP experiments, and OAT sensitivity analysis were conducted to identify further critical features affecting tensile strength. This study significantly improves the accuracy of polymer composite performance prediction. It advances the recognition and expression of graph features, providing actionable insights for optimizing material design and enhancing mechanical performance.

Imbalanced feature generation based on bootstrap power spectral curve for estimating respiratory rate

Rapid respiratory rate (RR) changes in older adults may indicate serious illness. Therefore, accurately estimating RR for cardiorespiratory fitness is essential. However, machine learning algorithm-related errors are unsuitable for medical decision-making processes because some data have a much larger sample size in the training set than in other sets. This difference in size refers to data imbalance. Therefore, we introduce a novel methodology combining bootstrap-based imbalanced feature generation (BIFG) with the Gaussian process for estimating RR and uncertainty, thereby addressing data imbalance. The sample difference between normal breathing (12–20 bpm), dyspnea (20 bpm), and hypopnea (<8 bpm) indicates significant data imbalance, which can affect the learning of the machine learning algorithm. Thus, the normal breathing part with much data is well-trained. The dyspnea and hypopnea parts with relatively little data are not well-trained, and this data imbalance causes significant errors concerning the reference variables in the actual dyspnea and hypopnea data parts. Hence, we use the parametric bootstrap model generated by artificial feature curves to estimate RR and solve this problem. As a result, the non-parametric bootstrap approach drastically increased the number of artificial feature curves. The generated artificial feature curves are selectively utilized for the highly imbalanced parts. Therefore, BIFG can be efficiently trained to predict the complex nonlinear relationships between the feature vectors obtained from the photoplethysmography signals and the reference RR. The proposed methodology exhibits more accurate predictive performance and uncertainty. The mean absolute errors are 0.89 and 1.44 beats per minute for RR using the proposed BIFG based on the two data sets.

Interpretable AI techniques unveil the factors and types of at-risk early warning: a case study in K-12 online learning

PurposeThis study focused on early warning in K-12 online education and aims to address the following research gaps: timing, model performance and understandability by completing the following analytic tasks: (1) determine the best prediction timing based on changes in course requirements; (2) compare early warning models with the correct evaluation indicators; (3) interpret a complex predictive model using interpretable AI techniques. Through holistic analyses, the case study shows that the semester can be divided into three stages with different course requirements. Students who failed to meet the course requirements of individual stages would be predicted as at-risk. Through interpretable AI techniques, the key predictors of individual stages were identified, and the factors causing a student to be predicted as at-risk were also revealed. In addition, multiple at-risk types were identified through the analyses.Design/methodology/approachThis study employed a variational autoencoder and time series segmentation to detect changes in learning behavioral patterns resulting from alterations in course requirements. The semester was divided into three stages. Subsequently, complex ensemble machine learning classifiers were employed for early warnings at each stage, enabling accurate prediction of students' learning performance. Interpretable AI techniques were employed to gain insight, identify characteristics of different at-risk types and suggest personalized interventions.FindingsThe case study analyzed 16,011 K-12 online school students. Major Findings include: (1) The semester was divided into three segments based on learning pattern transition points. XGBoost can identify the most at-risk students in each segment. Results indicate S2 is a relatively appropriate stage for early warning prediction and intervention. By the end of the second stage, the models can already identify over 75% of at-risk students. A student’s at-risk probability in previous stage has the most important influence on his performance in the next stage. If a student has high at-risk probability, he will have more possibility to be labeled as potential at-risk in the next stage, if he has low at-risk probability, he will have more possibility to be labeled as potential successful in the next stage, this reflects the influence between neighbor learning stages. We also found that, early warning can be most effectively conducted in the last week of each of these three stages. (2) From S1 to S3 stages, significant changes in the categories of key behaviors were observed, the key behavior at S1 stage was assignment viewing and submitting, at S2 stage was resource viewing and grade check, at S3 stage was assignment viewing and submitting, the discussion behavior became not important, while the testing behavior gradually became important, this reflected the changes in the course requirements and the learning strategy. It is evident that in specific stages of the semester, not all behaviors are better when more frequent; some behaviors, when excessive at the wrong time, can have a negative impact on predictive results. (3) Five types of at-risk students were found at each stage through interpretable AI analyses: high engaged at-risk, low engaged at-risk, testing at-risk, low interaction at-risk and un-persistent at-risk, which represented five types of learning patterns that conducted potential at-risk. With the progress of course, students adjusted learning strategy and might convert from one at-risk type to other at-risk type or successful type. The S2 stage was the most important stage, students had their last opportunity to reverse their at-risk trends at this stage. (4) Students who do not perform well in the early stages still have the opportunity to reverse the at-risk trend with effort in the later stages, emphasizing the importance of intervention at the end of the second stage. Similarly, students who perform well in the early stages can still face failure if they do not put in effort in the final stage. Low-engaged students tended to maintain their low-engaged status, resulting in failure at the end of the semester compared to other at-risk types, this was caused by that they had no learning activity to adjust their learning strategy.Research limitations/implicationsA follow-up study can focus on the development of personalized interventions and observe whether these interventions can influence potentially at-risk students' learning strategies, more effectively guiding students to align with course requirements and ultimately leading to improved performance.Originality/valueThis study is one of the few that focus on large-scale K-12 early warning prediction and aim to identify at-risk types via interpretable AI for personalized intervention. The authors found that the proposed method can effectively determine the best prediction timing, identify more at-risk students, and gain deep insight into students' learning processes. The authors confirm that the research in their work is original, and all the data given in the paper are real and authentic. The study has not been submitted to peer review and has not been accepted for publication in another journal.

Uncertainty quantification of compressive after impact strength in laminated composites using improved Bayesian ResNet deep learning model

AbstractThis study, for the first time, presented a novel Bayesian inference deep learning model based on a dilated convolutional ResNet framework, utilizing acoustic emission (AE) signals as inputs to quantify the uncertainty of the compression after impact (CAI) strength in composites. The output of the deep learning model included the mean CAI and the predictive aleatoric uncertainty. The epistemic uncertainty of the model was estimated by randomly sampling the input multiple times. AE signals were divided into a complete training dataset and a partially missing training dataset to investigate the impact of data insufficiency on the uncertainty of deep learning models. The results demonstrate that the deep learning model can provide highly accurate predictions of CAI performance. It is worth noting that on incomplete datasets, the epistemic uncertainty of the predictions increased significantly. To evaluate the effect of noise on the prediction of CAI strength, Gaussian noise with signal‐to‐noise ratios (SNRs) of 10, 5, and 2 dB was introduced during the prediction phase. The Bayesian posterior probability density function becomes lower and wider under the same impact energy as external noise increases. The model exhibited excellent regression performance under SNRs of 10 and 5 dB, validating the robustness of the proposed framework. As noise levels increased, the range of aleatoric uncertainty in the predictions expanded.Highlights A dilated convolution Bayesian deep learning framework for uncertainty quantification of compression after impact strength in laminated composites is proposed. The prediction and uncertainty quantification performance of the framework is evaluated. The effect of data missing and background noise on the uncertainty is investigated.

Review of Predictions of Impact Performances and Damages of Fiber Reinforced Composite using Machine Learning Approaches

Fiber-reinforced composites (FRCs) offer high specific mechanical properties like-exceptional strength, lightweight properties, and versatility, but their susceptibility to impact damage poses a significant challenge. Characterizing these while connecting processing methods, microstructure, and environmental factors to impact response has seen limited success through conventional modeling. But with increasing computational power and availability of data, machine learning techniques present opportunities in this domain enabling accurate prediction and robust monitoring of impact performance and damage. This review paper provides a comprehensive examination of the evolving landscape in predicting impact performances and damages of FRCs to optimize composites’ design through the lens of different supervised, unsupervised, blended, deep transfer learning and alternative approaches highlighting their strengths, limitations, and suitability for specific tasks. Methods encompassing artificial neural networks (ANNs), support vector machines (SVMs), and convolutional neural networks (CNNs) have exhibited promise in predicting performance and damage parameters respectively. Each section critically evaluates the strengths, limitations, and contributions of these approaches, providing a holistic view of their effectiveness.

Inside Cover Picture

Accurate prediction of chemical reaction performance is crucial for automated chemical synthesis. In this study, a novel multi‐modal chemical reaction prediction model is proposed, using graph and textual information, eliminating the need for computationally intensive DFT parameters with capability to handling reactions involving a fluctuating number of molecules. This model outperforms at least 7 generalized methods on 4 datasets, while maintaining chemical interpretability towards atomic and feature importance. Further details are comprehensively discussed in the article by Li et al. on pages 1230—1238. image

Parametric Sensitivity of a PEM Electrolyzer Mathematical Model: Experimental Validation on a Single-Cell Test Bench

Water electrolysis for hydrogen production is of great importance for the reliable use of renewable energy sources to have a clean environment. Electrolyzers play a key role in achieving the carbon-neutral target of 2050. Among the different types of water electrolyzers, proton exchange membrane water electrolyzers (PEMWEs) represent a well-developed technology that can be easily integrated into the smart grid for efficient energy management. In this study, a discrete dynamic mathematical model of a PEMWE was developed in MATLAB/Simulink to simulate cell performance under various operating conditions such as temperature, inlet flow rate, and current density loads. A lab-scale test bench was designed and set up, and a 5 cm2 PEMWE was tested at different temperatures (40–80 °C) and flow rates (3–12 mL/min), obtaining Linear Sweep Voltammetry (LSV), Cyclic Voltammetry (CV), Chrono-potentiometry (CP), and Electrochemical Impedance Spectroscopy (EIS) results for comparison and adjustment of the dynamic model. Sensitivity analysis of different operating variables confirmed that current density and temperature are the most influential factors affecting cell voltage. The parametric sensitivity of various chemical–physical and electrochemical parameters was also investigated. The most significant ones were estimated via non-linear least squares optimization to fine-tune the model. Additionally, strong correlations between these parameters and temperature were identified through regression analysis, enabling accurate performance prediction across the studied temperature range.

Abstract 4089: A novel NGS-based sample tracking system for robust sample chain of custody and quality control in fresh frozen and FFPE tumor samples

Abstract When processing large cohorts of samples, sample identity and functional quality are key to ensure the derived data is accurate and attributed to the appropriate sample. This is especially critical when working with precious and degraded material such as formalin-fixed paraffin-embedded (FFPE) samples. A potential sample swap or poor sample quality may result in incorrect data or delays due to re-processing. While traditional methods such as qPCR can be used for quality control, they may not be an effective predictor of sample success in Next-Generation Sequencing (NGS) based assays due to differential susceptibility to inhibitors and interfering substances. Performing a functional quality control assessment using the same methodology for data generation allows for a more accurate prediction of performance as it would be similarly affected by interfering substances or sample degradation. We have created a scalable, cost-effective sample tracking and functional quality control assay that may be assessed prior to performing more expensive downstream sequencing assays. This assay was designed to confirm consistency in sample identity during processing and confirm identity when multiple samples are submitted from the same subject. This amplicon panel targets autosomal SNPs with varying minor allele frequencies in different ethnic populations, limiting a bias towards variation in European ethnic populations. The average minor allele frequency for the panel developed for the general population is 0.45, more specifically 0.46 for European background, 0.44 for African background and 0.46 for Asian background based on data from the Allele Frequency Aggregator (ALFA) database. Most amplicons were designed to be less than 150 base pairs to increase the likelihood of success for samples with degraded DNA. Targets were included on the X and Y chromosome for sex determination. Tumor-normal paired samples were processed on the sample tracking assay, targeting regions throughout the exome to allow tracking of sample identity. When comparing “normal” samples extracted from blood to their paired FFPE tumor tissue, known matches had an average of 98.0% concordance, while known mismatches had an average of 40.4% concordance, with the highest known mismatch showing 61.9% concordance. Additionally, when comparing well-characterized samples to their corresponding data from the 1000 Genomes Project, samples had 99.4% concordance. This exemplifies a robust sample tracking assay that can ensure sample identity is correct. A strong quality control assay enables NGS laboratories to identify poor sample quality or degradation, allowing for re-extraction or new sample collection before processing on a more expensive assay. This newly developed sample tracking assay ensures both sample integrity and identity, leading to an overall increase in data quality for cancer research studies. Citation Format: Katherine Milligan, Yu Qiu, Andrea O'Hara, Carrie Ziemniak, David Corney, Haythem Latif. A novel NGS-based sample tracking system for robust sample chain of custody and quality control in fresh frozen and FFPE tumor samples [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 4089.

Performance and predictive modeling of surface carbonized plant fiber reinforced alkali-activated self-insulating mortar

Incorporating stalk fibers into alkali-activated cement faces challenges: alkaline exposure leaches sugars, while their smooth, low-energy surfaces weaken interfacial bonding. To address these issues, this study focuses on the surface carbonization of stalk fibers. The microstructural characteristics of surface-carbonized stalk were systematically investigated through multiple characterization methodologies. This research thoroughly examines the effects of carbonization temperature and duration on both mechanical properties and thermal transmission characteristics of alkali-activated cement-based composites, while conducting comparative analyses of different surface modification strategies for performance enhancement in stalk-reinforced systems. Experimental findings demonstrate that surface carbonization under standard curing conditions outperforms NaOH immersion treatment, with the resultant mortar exhibiting impressive mechanical properties surpassing those of untreated stalk composites. Notably, this composite material simultaneously achieves enhanced thermal insulation performance, manifesting a thermal conductivity coefficient of 0.2373 W/m2 K. Furthermore, a machine learning framework was developed to model the relationship between modified stalk microstructure and composite mechanical properties, enabling accurate prediction of material performance.

An improved GRU method for slope stress prediction

The stability of open-pit mine slopes is a complex nonlinear system. Stress variation is a significant influencing factor in the occurrence of landslide disasters and is also a key research focus in landslide early warning and risk assessment. However, traditional methods are confronted with challenges, including low prediction accuracy and poor robustness when dealing with nonlinear time series data. In order to address the aforementioned issues, the present paper proposes an intelligent prediction model based on Variational Mode Decomposition (VMD) and Dung Beetle Optimization (DBO), combined with an improved Gated Recurrent Unit (GRU), which is hereby referred to as the VMD-DBO-GRU-A model. The preliminary preprocessing of open pit mine slope stress data using VMD can provide high decomposition accuracy and can effectively extract localized features in the stress; The method introduces Dung Beetle Optimization (DBO) to determine the number of hidden neuron layers and the optimal learning rate for the GRU. This reduces the uncertainty of model parameters and minimizes the time required for parameter tuning; Self-attention mechanism is also added to assign different weights to the input features, which reduces the dependence on external information and is more adept at capturing the internal relevance of the data or features. In order to verify the validity of the model, experiments are conducted on a self-constructed stress dataset in this paper. The experimental results show that the root-mean-square error of the VMD-DBO-GRU-A model has decreased by 77% and 84% compared with the LSTM and SVM models, respectively, and the coefficient of determination is 0.9978, which fully verifies that the VMD-DBO-GRU-A model has an excellent comprehensive performance and high prediction accuracy, which is of great value for the practical application of landslide early warning for open-pit mines’ slopes.

Integrating machine learning and digital twin for strength prediction of CFRP/aluminum adhesive joints under hygrothermal conditions

AbstractThis study investigates the application of machine learning models integrated with a digital twin (DT) framework to predict and correlate the performance of carbon fibre‐reinforced polymer‐to‐aluminum adhesive joints subjected to hygrothermal aging. By combining experimental methods with machine learning techniques, the research aims to bridge the gap between the effects of natural and accelerated aging on adhesive joints. The joints were manufactured and then left to age naturally for a period of one to 3 years. For accelerated aging, the joints were subjected to hygrothermal conditions for a period of four to 50 days. Three‐point bending tests were utilized to evaluate the performance of the joints. To evaluate natural aging periods using accelerated aging data, five machine learning algorithms were used: support vector regression (SVR), artificial neural network (ANN), linear regression, random forest regression (RF) and XGBoost. scanning electron microscopy (SEM) analyses showed that moisture absorption caused a substantial change in the surface morphology of aluminum adherends, including increased roughness and crystalline formations. The results indicated that XGBoost has provided almost perfect predictions since MSE values equal to 0 were observed at all iterations, highlighting its accuracy and reliability. In contrast, the SVR and linear regression models demonstrated much lower accuracy, with clear differences observed in their predictions. The integration of digital twin with machine learning approaches turns out to be the most efficient method of real‐time adaptation of the model as well as accurate performance prediction, enhancing the durability and reliability of the composite structures.Highlights Strength prediction of adhesive joints by using Machine learning and digital twin. SEM revealed moisture‐induced changes in aluminum surface morphology. XGBoost model showed high prediction accuracy.

How distinct sources of nuisance variability in natural images and scenes limit human stereopsis.

Stimulus variability-a form of nuisance variability-is a primary source of perceptual uncertainty in everyday natural tasks. How do different properties of natural images and scenes contribute to this uncertainty? Using binocular disparity as a model system, we report a systematic investigation of how various forms of natural stimulus variability impact performance in a stereo-depth discrimination task. With stimuli sampled from a stereo-image database of real-world scenes having pixel-by-pixel ground-truth distance data, three human observers completed two closely related double-pass psychophysical experiments. In the two experiments, each human observer responded twice to ten thousand unique trials, in which twenty thousand unique stimuli were presented. New analytical methods reveal, from this data, the specific and nearly dissociable effects of two distinct sources of natural stimulus variability-variation in luminance-contrast patterns and variation in local-depth structure-on discrimination performance, as well as the relative importance of stimulus-driven-variability and internal-noise in determining performance limits. Between-observer analyses show that both stimulus-driven sources of uncertainty are responsible for a large proportion of total variance, have strikingly similar effects on different people, and-surprisingly-make stimulus-by-stimulus responses more predictable (not less). The consistency across observers raises the intriguing prospect that image-computable models can make reasonably accurate performance predictions in natural viewing. Overall, the findings provide a rich picture of stimulus factors that contribute to human perceptual performance in natural scenes. The approach should have broad application to other animal models and other sensory-perceptual tasks with natural or naturalistic stimuli.

Performance Variation of Transient Propulsion System Induced by Machining Errors in a Variable Combustion Chamber

Abstract Gun propulsion is a special mechanical system where machining errors in its components are inevitable. In particular, the gun barrel, which constitutes the internal combustion system, will be subject to structural changes caused by machining errors, leading to variations in propulsion performance. To comprehensively assess the reliability of the system, this study utilizes a coupled model that facilitates the simulation of the mechanical interaction between projectile and barrel, effectively capturing the system's nonlinear kinematic properties and the combustion behavior of propellant within the variable chamber. A classical simplified combustion model is used to improve computational efficiency without compromising the accuracy of critical performance predictions associated with the propulsion system. The model incorporates variations in the internal structure of the barrel due to machining errors and has been validated through comparison with experimental data. Based on the validated model, the performance variations of the propulsion system when it has machining errors are investigated. Additionally, the axial distribution of the core flow state parameters of propellant gas within the barrel is presented. The findings offer a profound understanding of the impact of structural precision on gun performance and provide valuable insights and recommendations for optimizing gun design.

Deep Transfer Learning-Based Performance Prediction Considering 3-D Flux in Outer Rotor Interior Permanent Magnet Synchronous Motors

Accurate performance prediction in the design phase of permanent magnet synchronous motors (PMSMs) is essential for optimizing efficiency and functionality. While 2-D finite element analysis (FEA) is commonly used due to its low computational cost, it overlooks important 3-D flux components such as axial leakage flux (ALF) and fringing flux (FF) that affect motor performance. Although 3-D FEA can account for these flux components, it is computationally expensive and impractical for rapid design iterations. To address this challenge, we propose a performance prediction method for interior permanent magnet synchronous motors (IPMSMs) that incorporates 3-D flux effects while reducing computational time. This method uses deep transfer learning (DTL) to transfer knowledge from a large 2-D FEA dataset to a smaller, computationally costly 3-D FEA dataset. The model is trained in 2-D FEA data and fine-tuned with 3-D FEA data to predict motor performance accurately, considering design variables such as stator diameter, axial length, and rotor design. The method is validated through 3-D FEA simulations and experimental testing, showing that it reduces computational time and accurately predicts motor characteristics compared to traditional 3-D FEA approaches.

Integrating a multitask graph neural network with DFT calculations for site-selectivity prediction of arenes and mechanistic knowledge generation

Abstract The accurate prediction of reaction performance based on empirical knowledge paves the way to efficient molecule design. Compared with the human-summarized reaction knowledge of a focal dataset, the machine-learned quantitative structure–performance relationship of larger-scale datasets is more effective at accessing the entire chemical space. Here we report a multitask learning workflow combined with a mechanism-informed graph neural network to predict site selectivity for ruthenium-catalysed C–H functionalization of arenes. The multitask architecture enables the acquisition of related knowledge from the simultaneous learning tasks. The embedded reaction graph bridges the gap between previous mechanistic studies and reaction representation. Along with this mechanistic embedding, the developed multitask model demonstrates excellent interpolative and extrapolative ability on the reported dataset composed of 256 reactions, achieving an average site-selectivity prediction accuracy of 0.934 with a standard deviation of 0.007. The prediction scope ranges from simple to fused arenes and was even extended to heterocyclic indole derivatives in the additional out of sample tests containing 14 unseen instances. Furthermore, interpretation of the model promotes the development of a para-selective mechanistic model verified by density functional theory calculations.

Interpretable Machine Learning for Evaluating Nanogenerators' Structural Design.

The limited battery life in modern mobile, wearable, and implantable electronics critically constrains their operational longevity and continuous use. Consequently, as a self-powered technology, triboelectric nanogenerators (TENGs) have emerged as a promising solution to this. Traditional approaches for evaluating TENG structural design typically require manual, repetitive, time-consuming, and high-cost finite element modeling or experiments. To overcome this bottleneck, we developed a fully automated platform that leverages machine learning (ML) techniques. Our framework contains an artificial neuron network-based surrogate model that can provide accurate and reliable performance predictions for any structural parameters and a TreeSHAP interpretable ML model that can generate precise global and local insights for TENG structural parameters. Our platform shows broad adaptability to multiple TENG structures. In summary, our platform is an integrated platform that utilizes interpretable ML techniques to solve the complex multidimensional TENG structural evaluation problem, marking a significant advancement in TENG design and supporting sustainable energy solutions in mobile electronics.