Deep Bayesian Slow Feature Extraction With Application to Industrial Inferential Modeling

Jiang, Chao; Lu, Yusheng; Zhong, Weimin; Huang, Biao; Song, Wenjiang; Tan, Dayu; Qian, Feng

doi:10.1109/tii.2021.3129888

“…Hence, Jiang et. al [30] proposed a deep Bayesian extension of the probabilistic slow feature analysis to enhance the nonlinear slow feature extraction using gated recurrent unit (GRU) neural networks. A summary of the proposed model is given below.…”

Section: Deep Bayesian Psfamentioning

confidence: 99%

“…The training dataset is used to train the network, and the validation dataset is utilized to obtain the optimal number of latent variables m, nodes in each layer, and each sample length l. The expectations in the lower bound equation are approximated using a Monte-Carlo estimate with twenty samples. The performance of the proposed method on the test data is compared with the other state-of-the-art techniques, including a regular GRU-based auto-encoder (GRU-AE), variational Bayesian complex PSFA (VBCPSFA) [21], and variable-wise deep Bayesian PSFA (VW-DBPSFA) [30]. Table I presents the latent variable dimension, observed variables reconstruction root mean square error (R-RMSE), target variable prediction root mean square error (P-RMSE), and the correlation between the prediction and the actual target variable (ρ) of different methods for two scenarios.…”

Section: A Simulation Case Studymentioning

confidence: 99%

“…Recent frameworks have considered variational learning of sequential data using recurrent neural networks. Jiang et al [30], in particular, considered the probabilistic slow feature analysis model as a prior distribution for the sequential data, thus facilitating the modelling preference. Although the proposed methodology can extract slow features from nonlinear sequential data, it has several shortcomings.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Nonlinear Semi-supervised Inference Networks for the Extraction of Slow Oscillating Features

Puli¹

2022

Preprint

Self Cite

0

View full text Add to dashboard Cite

<p>Due to physical and monetary constraints, quality-related variables in the process industries are generally difficult to measure using hardware sensors. Therefore, they are often not available as frequently as the other variables since obtaining them typically via laboratory analysis are time-consuming. On the other hand, the dynamic nature of easy-to-measure process data contains valuable predictive information about the quality variables. Consequently, modelling sequential data is beneficial in building a soft sensor that can predict the quality variable. Several techniques were proposed to extract dynamic features as measured data often suffers from noise and collinearity. This paper presents a semi-supervised oscillating slow feature inference network to address certain shortcomings of the existing solutions. The proposed learning algorithm uses a gated recurrent unit to learn a combined inference network for missing quality variable imputation and oscillating slow feature extraction from nonlinear process data. We evaluate the efficacy of the proposed methodology on a simulated and an industrial process.</p>

show abstract

“…As discussed in ( 17) -( 18), the existing deep Bayesian model for PSFA [30] assumes a different and simplified functional form that considers only the information from z z z k−1 and y y y k . The approximate posterior at time k is assumed to be Gaussian distribution, as shown below.…”

Section: B Inference Networkmentioning

confidence: 99%

Nonlinear Semi-supervised Inference Networks for the Extraction of Slow Oscillating Features

Puli¹,

Huang²

2022

Preprint

Self Cite

1

0

View full text Add to dashboard Cite

<p>Due to physical and monetary constraints, quality-related variables in the process industries are generally difficult to measure using hardware sensors. Therefore, they are often not available as frequently as the other variables since obtaining them typically via laboratory analysis are time-consuming. On the other hand, the dynamic nature of easy-to-measure process data contains valuable predictive information about the quality variables. Consequently, modelling sequential data is beneficial in building a soft sensor that can predict the quality variable. Several techniques were proposed to extract dynamic features as measured data often suffers from noise and collinearity. This paper presents a semi-supervised oscillating slow feature inference network to address certain shortcomings of the existing solutions. The proposed learning algorithm uses a gated recurrent unit to learn a combined inference network for missing quality variable imputation and oscillating slow feature extraction from nonlinear process data. We evaluate the efficacy of the proposed methodology on a simulated and an industrial process.</p>

show abstract

Multivariate long-time series traffic passenger flow prediction using causal convolutional sparse self-attention MTS-Informer

Liu,

Wang,

Hu

et al. 2023

Neural Comput & Applic

3

0

View full text Add to dashboard Cite

Deep Bayesian Slow Feature Extraction With Application to Industrial Inferential Modeling

Cited by 16 publications

References 28 publications

Nonlinear Semi-supervised Inference Networks for the Extraction of Slow Oscillating Features

Nonlinear Semi-supervised Inference Networks for the Extraction of Slow Oscillating Features

Nonlinear Semi-supervised Inference Networks for the Extraction of Slow Oscillating Features

Multivariate long-time series traffic passenger flow prediction using causal convolutional sparse self-attention MTS-Informer

Contact Info

Product

Resources

About