Calibrated Diffusion Tensor Estimation
Davood Karimi, Simon K. Warfield, Ali Gholipour

и другие.

arXiv (Cornell University), Год журнала: 2021, Номер unknown

Опубликована: Ноя. 21, 2021

It is highly desirable to know how uncertain a model's predictions are, especially for models that are complex and hard understand as in deep learning. Although there has been growing interest using learning methods diffusion-weighted MRI, prior works have not addressed the issue of model uncertainty. Here, we propose method estimate diffusion tensor compute estimation Data-dependent uncertainty computed directly by network learned via loss attenuation. Model Monte Carlo dropout. We also new evaluating quality predicted uncertainties. compare with standard least-squares bootstrap-based computation techniques. Our experiments show when number measurements small more accurate its better calibrated than methods. uncertainties can highlight biases, detect domain shift, reflect strength noise measurements. study shows importance practical value modeling prediction learning-based MRI analysis.

Язык: Английский

Challenges in data-driven geospatial modeling for environmental research and practice DOI Creative Commons
Diana Koldasbayeva, Polina Tregubova, Mikhail Gasanov

и другие.

Nature Communications, Год журнала: 2024, Номер 15(1)

Опубликована: Дек. 19, 2024

Machine learning-based geospatial applications offer unique opportunities for environmental monitoring due to domains and scales adaptability computational efficiency. However, the specificity of data introduces biases in straightforward implementations. We identify a streamlined pipeline enhance model accuracy, addressing issues like imbalanced data, spatial autocorrelation, prediction errors, nuances generalization uncertainty estimation. examine tools techniques overcoming these obstacles provide insights into future AI developments. A big picture field is completed from advances processing general, including demands industry-related solutions relevant outcomes applied sciences. In this scoping review, authors explore challenges implementing data-driven models—namely machine learning deep algorithms—in research.

Язык: Английский

Процитировано

9
Heat Recovery Integration in a Hybrid Geothermal-based System Producing Power and Heating Using Machine Learning Approach to Maximize Outputs DOI Creative Commons

Hatem N.E. Gasmi,

Azher M. Abed, Ashit Kumar Dutta

и другие.

Case Studies in Thermal Engineering, Год журнала: 2024, Номер unknown, С. 105210 - 105210

Опубликована: Сен. 1, 2024

Язык: Английский

Процитировано

4
Informer–SVR: Traffic Volume Prediction Hybrid Model Considering Residual Autoregression Correction DOI
Chang Xu, Yichen Chen, Qingwei Zeng

и другие.

Journal of Transportation Engineering Part A Systems, Год журнала: 2025, Номер 151(4)

Опубликована: Янв. 23, 2025

Язык: Английский

Процитировано

0
Hybrid Semi-mechanistic and Machine Learning Solubility Regression Modeling for Crystallization Process Development DOI
Gustavo Lunardon Quilló, Satyajeet Bhonsale, A. Collas

и другие.

Crystal Growth & Design, Год журнала: 2025, Номер unknown

Опубликована: Фев. 10, 2025

Solubility regression modeling is foundational for several chemical engineering applications, particularly crystallization process development. Traditionally, these models rely on parametric semimechanistic approaches such as the Van't Hoff Jouyban-Acree (VH-JA) cosolvency model. Although generally provide narrow prediction intervals, they can exhibit increased bias when dealing with significant solute heat capacities or complex mixture effects. This study explores machine learning, including Random Forests, Support Vector Machines, Gaussian Process Regression, and Neural Networks, potential alternatives. While most learning offered a lower training error, it was observed that their predictive quality quickly deteriorates further from data. Hence, hybrid approach explored to leverage low of variance VH-JA model through heterogeneous locally weighted bagging ensembles. Key methodology quantifying, tracking, minimizing uncertainty using ensemble. illustrated case solubility ketoconazole in binary mixtures 2-propanol water. The optimal ensemble, comprising 58% stepwise 42% models, reduced root-mean-squared error maximum absolute percentage by ≈30% compared full VH-JA, while preserving comparable interval.

Язык: Английский

Процитировано

0
Optimization Design of Protective Helmet Structure Guided by Machine Learning DOI Open Access
Yongxing Chen, Junlong Wang, Long Peng

и другие.

Processes, Год журнала: 2025, Номер 13(3), С. 877 - 877

Опубликована: Март 16, 2025

With increasing digitization worldwide, machine learning has become a crucial tool in industrial design. This study proposes novel learning-guided optimization approach for enhancing the structural design of protective helmets. The optimal model was developed using algorithms, including random forest (RF), support vector (SVM), eXtreme gradient boosting (XGB), and multilayer perceptron (MLP). hyperparameters these models were determined by ten-fold cross-validation grid search. experimental results showed that RF had best predictive performance, providing reliable framework guiding optimization. SHapley Additive exPlanations (SHAP) method on contribution input features show three structures—the transverse curvature at foremost point forehead, helmet forehead bottom edge elevation angle, maximum along longitudinal centerline forehead—have highest both goals. research achievement provides an objective helmets, further promoting development

Язык: Английский

Процитировано

0
Enhancing Brain Age Estimation Under Uncertainty: A Spectral-normalized Neural Gaussian Process Approach Utilizing 2.5D Slicing DOI Creative Commons
Zeqiang Linli,

Xingcheng Liang,

Zhenhua Zhang

и другие.

NeuroImage, Год журнала: 2025, Номер unknown, С. 121184 - 121184

Опубликована: Апрель 1, 2025

Brain age gap, the difference between estimated brain and chronological via magnetic resonance imaging, has emerged as a pivotal biomarker in detection of abnormalities. While deep learning is accurate estimating age, absence uncertainty estimation may pose risks clinical use. Moreover, current 3D models are intricate, using 2D slices hinders comprehensive dimensional data integration. Here, we introduced Spectral-normalized Neural Gaussian Process (SNGP) accompanied by 2.5D slice approach for seamless integration single network with low computational expenses, extra without added model complexity. Subsequently, compared different methods Pearson correlation coefficient, metric that helps circumvent systematic underestimation during training. SNGP shows excellent generalization on dataset 11 public datasets (N=6327), competitive predictive performance (MAE=2.95). Besides, demonstrates superior (MAE=3.47) an independent validation set (N=301). Additionally, conducted five controlled experiments to validate our method. Firstly, adjustment improved accelerated aging adolescents ADHD, 38% increase effect size after adjustment. Secondly, exhibited OOD capabilities, showing significant differences across Asian non-Asian datasets. Thirdly, DenseNet backbone was slightly better than ResNeXt, attributed DenseNet's feature reuse capability, robust set. Fourthly, site harmonization led decline performance, consistent previous studies. Finally, significantly outperformed methods, improving increasing In conclusion, present cost-effective method uncertainty, utilizing slicing enhanced showcasing promise applications.

Язык: Английский

Процитировано

0
Rapid changes in terrestrial carbon dioxide uptake captured in near-real time from a geostationary satellite: The ALIVE framework DOI Creative Commons
Danielle Losos, Sadegh Ranjbar, Sophie Hoffman

и другие.

Remote Sensing of Environment, Год журнала: 2025, Номер 324, С. 114759 - 114759

Опубликована: Апрель 14, 2025

Язык: Английский

Процитировано

0
AI-Driven Solutions for Regression Testing: Insights from Bangladesh Software Industry DOI
Md Mahbub Alam,

Sabrina Islam Priti,

Selim Ahmed

и другие.

Smart innovation, systems and technologies, Год журнала: 2025, Номер unknown, С. 495 - 508

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

0
Composite quantile regression with XGBoost using the novel arctan pinball loss DOI Creative Commons
Laurens Sluijterman, Frank Kreuwel, Eric Cator

и другие.

International Journal of Machine Learning and Cybernetics, Год журнала: 2025, Номер unknown

Опубликована: Май 20, 2025

Язык: Английский

Процитировано

0
Recurrence Multilinear Regression Technique for Improving Accuracy of Energy Prediction in Power Systems DOI Creative Commons
Quota Alief Sias, Rahma Gantassi, Yonghoon Choi

и другие.

Energies, Год журнала: 2024, Номер 17(20), С. 5186 - 5186

Опубликована: Окт. 18, 2024

This paper demonstrates how artificial intelligence can be implemented in order to predict the energy needs of daily households using both multilinear regression (MLR) and single linear (SLR) methods. As a basic implementation, SLR makes use one input variable, which is total amount generated as an input. The MLR implementation involves multiple variables being taken from various sources, including gas, coal, geothermal, wind, water, biomass, oil, etc. All these are derived detailed production data sources. purpose this demonstrate that it possible analyze demand supply directly together way produce more in-depth analysis. By analyzing previous periods time, prediction made. Compared found perform better because able achieve smaller error value. Furthermore, forecasting pattern carried out sequentially based on periodic pattern, so calls method recurrence (RMLR) method. also creates pre-clustering K-Means algorithm before improve accuracy. Other models such exponential GPR, sequential XGBoost, seq2seq LSTM used for comparison. results evaluated by calculating MAE, RMSE, MAPE, MAPA, time execution all models. simulation show fastest best model obtains smallest (3.4%) RMLR clustered weekly period.

Язык: Английский

Процитировано

2