Broad Learning System with Locality Sensitive Discriminant Analysis for Hyperspectral Image Classification

In this paper, we propose a new method for hyperspectral images (HSI) classification, aiming to take advantage of both manifold learning-based feature extraction and neural networks by stacking layers applying locality sensitive discriminant analysis (LSDA) to broad learning system (BLS). BLS has been proven to be a successful model for various machine learning tasks due to its high feature representative capacity introduced by numerous randomly mapped features. However, it also produces redundancy, which is indiscriminate and finally lowers its performance and causes heavy computing demand, especially in cases of the input data bearing high dimensionality. In our work, a manifold learning method is integrated into the BLS by inserting two LSDA layers before the input layer and output layer separate, so the spectral-spatial HSI features are fully utilized to acquire the state-of-the-art classification accuracy. The extensive experiments have shown our method’s superiority.

Elucidation of ball bearing performance utilizing product functions of vibration signals and locality sensitive discriminant analysis

Dimensionality is a decent term utilized for the dimension-related issues of feature vectors for the condition assessment of bearings. It has been a challenging task while dealing with the characteristic information conservation of the sampled data. Feature vectors with higher dimensions provide an accurate description of the condition, while lower dimension vectors are easy to be classified. These conditions make the dimensionality of feature space a big challenge in the performance evaluation of critical machine parts like bearings. This paper presents a judicious and wise application of a linear approach, locality sensitive discriminant analysis along with the rational use of local mean decomposition in performance degradation assessment. The combination effectively solves the problems of insufficient training samples and large dimensionality of fault features, which imparts excessive noise and causes loss of competent hidden characteristics. In such cases, local structures of feature space are more crucial than the global one. These obstacles can be answered by imparting locality sensitive discriminant analysis as an initial operation. Locality sensitive discriminant analysis is a linear dimensionality reduction tool, which explores the precise projections that amplify the margin between data points and prepare a conservation aid to preserve faulty information of bearings. The steps are followed to achieve the same: decomposition of vibration signal into product functions; calculation of fault features; third, higher dimensionality of the features is reduced with the implementation of locality sensitive discriminant analysis and some prime features are selected with the proposed criterion; the reduced features are further clustered and a trained model of assessment is prepared. Finally, health indicator is calculated from the trained model and test features. The proposed technique is verified on two bearing datasets. The superiority of the technique has been envisaged by comparing the method with different similar assessment methods i.e. time domain features, linear discriminant analysis, and principal component analysis.