Gaussian process regression: active data selection and test point rejection

Seo, Sambu; Wallat, Marko; Graepel, Thore; Obermayer, Klaus

doi:10.1109/ijcnn.2000.861310

Cited by 34 publications

(7 citation statements)

References 3 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The approaches presented therein are based on variance reduction using Bayesian models, like Gaussian process regression (GPR), cf. Seo et al (2000). Such Bayesian approaches can have a tremendous advantage when working with experimental measurements.…”

Section: Related Workmentioning

confidence: 99%

Efficient Exploration of Microstructure-Property Spaces via Active Learning

et al. 2022

View full text Add to dashboard Cite

In materials design, supervised learning plays an important role for optimization and inverse modeling of microstructure-property relations. To successfully apply supervised learning models, it is essential to train them on suitable data. Here, suitable means that the data covers the microstructure and property space sufficiently and, especially for optimization and inverse modeling, that the property space is explored broadly. For virtual materials design, typically data is generated by numerical simulations, which implies that data pairs can be sampled on demand at arbitrary locations in microstructure space. However, exploring the space of properties remains challenging. To tackle this problem, interactive learning techniques known as active learning can be applied. The present work is the first that investigates the applicability of the active learning strategy query-by-committee for an efficient property space exploration. Furthermore, an extension to active learning strategies is described, which prevents from exploring regions with properties out of scope (i.e., properties that are physically not meaningful or not reachable by manufacturing processes).

show abstract

Section: Related Workmentioning

confidence: 99%

Efficient Exploration of Microstructure-Property Spaces via Active Learning

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The GPR methodology is based a Gaussian model for the training data error , which leads to a Gaussian likelihood function for the regressors and, together with a Gaussian assumption for the prior distribution of the model parameters, it leads to a model for the prediction that includes a mean and a variance of the prediction. This is advantageous over other regression approaches because provided the likelihood and the prior assumptions are correct, they provide not only a prediction but also a confidence interval over this prediction that allows to determine whether the quality of this prediction is acceptable or not for the application at hand [5]. Also, the nature of GPR is such that it does not have free parameters, so no cross validation is needed in the training process.…”

Section: Introductionmentioning

confidence: 99%

Probabilistic analysis of solar cell optical performance using Gaussian processes

Jaiswal,

Martínez-Ramón,

Busani

2021

Preprint

View full text Add to dashboard Cite

This work investigates application of different machine learning based prediction methodologies to estimate the performance of silicon based textured cells. Concept of confidence bound regions is introduced and advantages of this concept are discussed in detail. Results show that reflection profiles and depth dependent optical generation profiles can be accurately estimated using gaussian processes with exact knowledge of uncertainty in the prediction values.It is also shown that cell design parameters can be estimated for a desired performance metric.

show abstract

“…Therefore it is blind with parameter adjustment about kernel function, and it is limited to learn the noise model, and also it is difficult to give the degree of influence on the model. In this paper, Gaussian process method is used for gearbox condition monitoring [9][10][11][12], this method is able to conduct an independent analysis of signal and noise, and give the accurate impact of noise on the model. This paper makes an analysis on SCADA data of 1.5MW unit, establishes the gearbox temperature model and predicts the temperature of gearbox.…”

Section: Introductionmentioning

confidence: 99%

Wind turbine gearbox forecast using Gaussian process model

Wang

Zhou

Guo

2014

The 26th Chinese Control and Decision Conference (2014 CCDC)

View full text Add to dashboard Cite

For wind farms, wind turbine condition monitoring is important to reduce maintenance costs and improve the competitiveness in the electricity market, particularly for offshore wind farms. This paper seeks to establish wind turbine gearbox temperature model under the normal working state using Gaussian process, the forecast and evaluation of temperature is also described. Within the Bayesian context, the paper aims to training Gaussian process, using the maximum likelihood optimized approach to find the optimal hyperparameters. For large-scale regression tasks, a novel method using Cholesky decomposition to avoid ill-conditioned matrix is described. Another method using matrix caching to speed up the inverse of matrix calculation is proposed. In addition, the optimized Gaussian model is used to predict the gearbox validation data and compare with SVM (support vector machine) and BPNN (neural network) this two methods. By comparing the simulation results, Gaussian process gearbox temperature model demonstrates higher prediction accuracy. The model is a valuable object for condition monitoring.

show abstract

Gaussian process regression: active data selection and test point rejection

Cited by 34 publications

References 3 publications

Efficient Exploration of Microstructure-Property Spaces via Active Learning

Efficient Exploration of Microstructure-Property Spaces via Active Learning

Probabilistic analysis of solar cell optical performance using Gaussian processes

Wind turbine gearbox forecast using Gaussian process model

Contact Info

Product

Resources

About