Abstract:Given that there exists input uncertainty caused by the noise embedded in industrial data, this study proposes a relevance vector machine (RVM) prediction model with input noise. Due to the fact that the marginal likelihood cannot be analytically calculated when introducing the input uncertainty, a Gaussian approximation is proposed in this study on the basis of the law of total expectation and the law of total covariance. Furthermore, to approximate the posterior distribution over the model weights, this study employs the Markov chain Monte Carlo algorithm, where a Gaussian proposal distribution is designed to draw new samples. In the prediction stage, a Gaussian approximation is also designed for a new testing input in order for the input uncertainty to be reflected in the estimation of output variance. To verify the effectiveness of the proposed method, four synthetic data sets, four benchmark data sets, and two industrial data sets are employed in the comparative experiments. The results indicate that the proposed RVM with uncertain input outperforms other approaches, and it also performs better on the time series prediction issue.

What problem does this paper attempt to address?

The paper primarily addresses the issue of input uncertainty (caused by noise) present in industrial data by proposing an improved Relevance Vector Machine (RVM) model, namely the Relevance Vector Machine with Uncertain Inputs (RVMUI). The core contributions of the paper can be summarized as follows: 1. **Problem Background and Challenges**: - In complex industrial production environments, due to the complexity of the production process, the collected data usually contains noise, leading to input uncertainty in the model. - Existing methods such as Artificial Neural Networks and Support Vector Machines, although capable of handling such data, often overlook the impact of input noise. 2. **Proposed Solution**: - A new RVMUI model is proposed, which considers the noise in the input data and incorporates it into the prediction model. - To address the issue of the marginal likelihood function being uncomputable due to the introduction of input uncertainty, the paper proposes a Gaussian approximation method. - The Markov Chain Monte Carlo (MCMC) algorithm is used to approximate the posterior distribution, and a Gaussian proposal distribution is designed to generate new samples. - In the prediction phase, a Gaussian approximation is also designed to reflect the impact of input uncertainty of new test inputs on the output variance estimation. 3. **Experimental Validation**: - Comparative experiments were conducted on four synthetic datasets, four benchmark datasets, and two industrial datasets to verify the effectiveness of the proposed method. - The experimental results show that the proposed RVMUI model outperforms other methods in handling data with input noise and demonstrates better performance in time series prediction tasks. 4. **Technical Details**: - A Gaussian approximation based on the law of total expectation and the law of total covariance is used to handle the marginal likelihood function. - The Metropolis-Hastings algorithm is utilized for sampling to approximate the complex posterior distribution. - The predictive distribution is also approximated using a similar method, thereby propagating input uncertainty into the predictive variance. In summary, this paper aims to effectively address the issue of input uncertainty in industrial data by improving the RVM model and validates the effectiveness and superiority of the proposed method through a series of experiments.

A Gaussian Approximation of Marginal Likelihood in Relevance Vector Machine for Industrial Data With Input Noise

Noise model based ν-support vector regression with its application to short-term wind speed forecasting

A Novel Bayesian Robust Model and Its Application for Fault Detection and Automatic Supervision of Nonlinear Process

Bayesian Nonlinear Gaussian Mixture Regression and Its Application to Virtual Sensing for Multimode Industrial Processes.

Combined modeling of sparse and dense noise for improvement of Relevance Vector Machine

A Self-Optimal Relevance Vector Machine with Multiple-Attribute Gaussian Kernel Functions

A Novel Incremental Gaussian Mixture Regression and Its Application for Time-varying Multimodal Process Quality Prediction

Semisupervised Robust Modeling of Multimode Industrial Processes for Quality Variable Prediction Based on Student's T Mixture Model.

Dynamic Probabilistic Latent Variable Model for Process Data Modeling and Regression Application

Dynamic Variational Bayesian Student's T Mixture Regression With Hidden Variables Propagation for Industrial Inferential Sensor Development

A Random Field Model and its Application in Industrial Production

Nonlinear Soft Sensor Development Based on Relevance Vector Machine

Non-Gaussian Industrial Process Monitoring with Probabilistic Independent Component Analysis

Investigation on the Construction of the Relevance Vector Machine Based on Cross Entropy Minimization

A Probabilistic Quality-Relevant Monitoring Method with Gaussian Mixture Model

Fuzzy Relevance Vector Machine for Learning from Unbalanced Data and Noise

A Comparative Study of Relevant Vector Machine and Support Vector Machine in Uncertainty Analysis

Adaptive relevance vector machine combined with Markov-chain-based importance sampling for reliability analysis

A new multikernel relevance vector machine based on the HPSOGWO algorithm for predicting and controlling blast-induced ground vibration

Heteroscedastic Relevance Vector Machine

Noise-Aware Sparse Gaussian Processes and Application to Reliable Industrial Machinery Health Monitoring