Outliers In Training Set Research Articles

Kronecker-decomposable robust probabilistic tensor discriminant analysis

As a generative model, probabilistic linear discriminant analysis (PLDA) has achieved good performance in supervised learning tasks. The model incorporates both within-individual and between-individual variation, and remaining unexplained data variation is assumed to follow Gaussian distribution. However, the assumption of Gaussian distribution makes the model sensitive to the presence of noise and outliers in training set. To address this issue, this paper proposes a robust probabilistic linear discriminant analysis model by assuming Laplace prior on the noise term. Instead of solving high-dimensional linear systems, we embed a Kronecker-decomposable component in the new model for tensor data, significantly reducing the size of problems. As the non-conjugacy of Laplace distribution complicates the calculation of the posteriors of latent variables, we express it to a hierarchical architecture using an Inverse Gamma distribution and then adopt variational expectation–maximization (EM) algorithm to learn model parameters. The reconstruction and classification experiments on several public databases show the superiority of the proposed model compared with the state-of-the-art LDA-based algorithms.

Robust least squares one-class support vector machine

Abstract In comparison with the conventional one-class support vector machine (OCSVM), least squares OCSVM (LS-OCSVM) can describe similarity between a new-coming sample and training set more accurately. However, LS-OCSVM is very sensitive to outliers in training set. The main reason lies that the values of square error function for outliers are relatively large, which makes LS-OCSVM put more emphasis on these outliers. To enhance the robustness of LS-OCSVM against outliers, a novel robust LS-OCSVM based on correntropy loss function is proposed. As a result, the unbounded convex square loss function of LS-OCSVM is substituted by a bounded nonconvex correntropy loss function. Experimental results on synthetic and benchmark data sets show that robust LS-OCSVM possesses better anti-outlier and generalization abilities in comparison with its related approaches.

Bearing performance degradation assessment based on lifting wavelet packet decomposition and fuzzy c-means

Bearing performance degradation assessment is crucial to realize condition-based maintenance. In this paper, a new method for it is proposed based on lifting wavelet packet decomposition and fuzzy c-means. Feature vectors are composed of the node energies of lifting wavelet packet decomposition. Normal and final failure data are used as training samples to build assessment model utilizing fuzzy c-means, and the subjection of tested data to normal state is defined as the degradation indicator, which has intuitionistics explanation related to degradation degree. Results of its application to accelerated bearing life test show that this indicator can reflect effectively performance degradation of bearing. And after discussing the influence of outliers in training set, a robust strategy is proposed.