Adaptive parameter estimation of GMM and its application in clustering

Zhao, Linchang; Shang, Zhaowei; Tan, Jin; Luo, Xiaoliu; Zhang, Taiping; Wang, Yu; Tang, Yuan Yan

doi:10.1016/j.future.2020.01.012

Cited by 8 publications

(5 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…For the baseline group, there are three populations in the baseline experiment, and the populations are separable. In this case, the subpopulations can be estimated by Bayesian variational inference [62][63]31]. For the controlled group 1 and 2, the optimization method is suitable for the subpopulation model to achieve the optimal evaluation criteria [64].…”

Section: Results and Analysismentioning

confidence: 99%

Data analytics and optimization for smart industry

2020

View full text Add to dashboard Cite

Hit rate is a key performance metric in predicting process product quality in integrated industrial processes. It represents the percentage of products accepted by downstream processes within a controlled range of quality. However, optimizing hit rate is a non-convex and challenging problem.To address this issue, we propose a data-driven quasi-convex approach that combines factorial hidden Markov models, multitask elastic net, and quasi-convex optimization. Our approach converts the original non-convex problem into a set of convex feasible problems, achieving an optimal hit rate. We verify the convex optimization property and quasi-convex frontier through Monte Carlo simulations and real-world experiments in steel production. Results demonstrate that our approach outperforms classical models, improving hit rates by at least 41.11% and 31.01% on two real datasets. Furthermore, the quasi-convex frontier provides a reference explanation and visualization for the deterioration of solutions obtained by conventional models.

show abstract

Section: Results and Analysismentioning

confidence: 99%

Data analytics and optimization for smart industry

2020

View full text Add to dashboard Cite

show abstract

“…Depending on the data, some clustering methods are more suitable than others. Depending on the expected shape of the clusters, it will be, for example, more appropriate to use K-means (circular) or GMM (ellipsoidal) [20], or for any other reason, another clustering method. In our work, we tried with AP or GMM instead of K-means in the first level clustering.…”

Section: Discussionmentioning

confidence: 99%

A Multilevel Clustering Method for Risky Areas in the Context of Avalanche Danger Management

Pagnier

Pourraz

Coquin

et al. 2022

Information Processing and Management of Uncertainty in Knowledge-Based Systems

View full text Add to dashboard Cite

In the context of avalanche risk management, we study the spatial variability of rainfall conditions, which is one of the main parameters that induce natural avalanches. This paper focuses on the geographical variability of the snow overload due to recent precipitations. We propose a generic approach applicable at a larger scale and, for this reason, without relying on any expert knowledge. Our proposal is a multilevel clustering process based on classical methods processed in sequence to take advantage of each one. As a result, the multilevel clustering process outputs four main detected weather trends that affect the French Alps. The developed process is generic enough to be used in other areas. Our work is intended to positively impact and improve the current and future decision support methods and tools for mountain practitioners.

show abstract

“…[ 44 ] Gao et al [ 45 ] proposed an error compensation method based on GMM in order to reduce the influence of model parameter uncertainty in the oil and gas production process. Zhao et al [ 46 ] proposed a novel parameter estimation algorithm that combines the Tsallis entropy and a deterministic annealing (DA) algorithm on the basis of the variational Bayesian expectation–maximization (VBEM). And it implements the parameter estimation and selects the optimal components of GMM simultaneously.…”

Section: Introductionmentioning

confidence: 99%

Offline identification and output prediction for a class of SISO Wiener process

et al. 2022

View full text Add to dashboard Cite

In this paper, a simplified Wiener structure (SWS) for single‐input‐single‐output (SISO) Wiener processes and an identification method based on Gaussian mixture model (GMM) and expectation maximization (EM) algorithm are proposed. The vast majority of industrial processes can be regarded as approximately monotonic non‐linear processes. Approximately, a monotonic characteristic is introduced into the non‐linear module and a rich dynamic characteristic is added into the linear module of the Wiener model. Thus, SWS is exploited and has strong generalization ability. Because GMM can describe the arbitrary distribution of sample data in theory, it is used to accurately describe the output data, including the system error term of SWS. Hence, a statistical model (Equation (17)) is obtained. Then, the EM algorithm is introduced to identify the parameters of the statistical model that contains the parameters of SWS. In the end, two numerical examples demonstrate the effectiveness of both the SWS and the GMM‐EM‐based iterative offline identification algorithm.

show abstract

Adaptive parameter estimation of GMM and its application in clustering

Cited by 8 publications

References 22 publications

Data analytics and optimization for smart industry

Data analytics and optimization for smart industry

A Multilevel Clustering Method for Risky Areas in the Context of Avalanche Danger Management

Offline identification and output prediction for a class of SISO Wiener process

Contact Info

Product

Resources

About