Multi-task and multi-kernel Gaussian process dynamical systems

Korkinof, Dimitrios; Demiris, Yiannis

doi:10.1016/j.patcog.2016.12.014

Cited by 7 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In contrast to MTGPs, an MTDS allows much greater flexibility of the dynamic variation. The MT GP dynamical system of Korkinof & Demiris (2017) mitigates some of these limitations, but it retains a simple linear combination of latent dynamics with a Gaussian density over the combination.…”

Section: Related Workmentioning

confidence: 99%

Customizing Sequence Generation with Multi-Task Dynamical Systems

Bird,

Williams

2019

Preprint

View full text Add to dashboard Cite

Dynamical system models (including RNNs) often lack the ability to adapt the sequence generation or prediction to a given context, limiting their real-world application. In this paper we show that hierarchical multi-task dynamical systems (MTDSs) provide direct user control over sequence generation, via use of a latent code z that specifies the customization to the individual data sequence. This enables style transfer, interpolation and morphing within generated sequences. We show the MTDS can improve predictions via latent code interpolation, and avoid the long-term performance degradation of standard RNN approaches.

show abstract

Section: Related Workmentioning

confidence: 99%

Customizing Sequence Generation with Multi-Task Dynamical Systems

Bird,

Williams

2019

Preprint

View full text Add to dashboard Cite

show abstract

“…Besides, in GPR, values of the optimal model parameter (hyper-parameter) can be determined by the maximum likelihood estimation of the training samples. However, it is commonly suffered by being sensitive to the initial value and easily falling into the local extremum (Korkinof and Demiris, 2017). Especially in multi-kernel GPR, the accurate hyperparameter values are hard to get, because the number of hyper-parameters will increase greatly.…”

Section: Introductionmentioning

confidence: 99%

Soft sensor for the moisture content of crude oil based on multi-kernel Gaussian process regression optimized by an adaptive variable population fruit fly optimization algorithm

Han

et al. 2019

Transactions of the Institute of Measurement and Control

View full text Add to dashboard Cite

In the practical oilfield production, it has great significance to realize timely and accurate measurement of the moisture content of crude oil. However, there are some drawbacks in the traditional measurement methods, such as: non-real time, high cost, labor-consume, vulnerability to environmental impacts, and so on. In order to solve these problems, a soft sensor model based on multi-kernel Gaussian process regression optimized by an adaptive variable population fruit fly optimization algorithm (APFOA-MKGPR) is presented in this paper. A multiple kernels-based Gaussian process regression method is utilized to deal with the practical production process characterised by multiple operating phases, noises, strong nonlinearity and dynamic. In the multi-kernel function, many parameters (five hyper-parameters in the multi-kernel function and three weights of each kernel function) need to be accurately given, which is difficult to be effectively optimized by the maximum likelihood estimation. So, a swarm intelligence-based adaptive variable population fruit fly optimization algorithm (APFOA) is proposed to train the best model parameters. A novel adaptive variable population mechanism is developed to adaptively adjust the population size and the random flight distance during the iterations, which can realize a combination of the global searching and the local searching for the optimal solutions. The proposed method is verified by four benchmark functions and the actual production data of one oil well, and experimental results show the effectiveness for accurate prediction of the moisture content of crude oil.

show abstract

“…by periodic kernels. VGPDSs were also applied in many fields, such as phoneme classification [19], video repairing [20] and multi-task motion modeling [21]. The VDM-GPDS considers the dependence of multiple outputs and introduces convolution processes to explicitly depict multi-output dependence.…”

Section: Introductionmentioning

confidence: 99%

A Variant of Gaussian Process Dynamical Systems

Zhao,

Fei,

Sun

2019

Preprint

View full text Add to dashboard Cite

In order to better model high-dimensional sequential data, we propose a collaborative multi-output Gaussian process dynamical system (CGPDS), which is a novel variant of GPDSs. The proposed model assumes that the output on each dimension is controlled by a shared global latent process and a private local latent process. Thus, the dependence among different dimensions of the sequences can be captured, and the unique characteristics of each dimension of the sequences can be maintained. For training models and making prediction, we introduce inducing points and adopt stochastic variational inference methods.

show abstract

Multi-task and multi-kernel Gaussian process dynamical systems

Cited by 7 publications

References 31 publications

Customizing Sequence Generation with Multi-Task Dynamical Systems

Customizing Sequence Generation with Multi-Task Dynamical Systems

Soft sensor for the moisture content of crude oil based on multi-kernel Gaussian process regression optimized by an adaptive variable population fruit fly optimization algorithm

A Variant of Gaussian Process Dynamical Systems

Contact Info

Product

Resources

About