20 April 2021 in IEEE Transactions on Industrial Informatics
Data in modern industrial applications and data science present multidimensional progressively, the dimension and the structural complexity of these data are becoming extremely high, which renders existing data analysis methods and machine learning algorithms inadequate to the extent. In addition, high-dimensional data in actual scenarios often share some common latent components and patterns, it is necessary and significant to analyze such data in an associative manner, rather than treating them independently. Considering the problem of data islands and data privacy that is prevalent in the industry. In this article, we propose the first joint high-order orthogonal iterative (J-HOOI) algorithm for simultaneous tensor decomposition and federated tensor decomposition (FTD) model for feature extraction and dimension reduction of high-dimensional industrial data under the federated learning framework. Moreover, we also develop a secure federated computation process based on the J-HOOI method. Using this method, multiple participants iteratively calculate the local factor matrices and transfer the local information to the parameter server, which aggregates the local information to generate the globally updated factor matrices. Finally, each client generates globally compressed features by projecting local data onto these common potential spaces. We have demonstrated with real-world industrial datasets that our approach is similar to a centralized training model in decomposition accuracy and classification accuracy while respecting privacy.
12 October 2021 in IEEE Transactions on Network Science and Engineering
Multiple clusterings are conducive to discovering different data patterns hidden in data from different perspectives, so it has tremendous value in applications like community detection, resource recommendation, and gene expression, etc. To solve the problem that the existing multiple clustering approaches are mainly oriented to low-dimensional single-domain data and are not suitable for big data in Cyber-Physical-Social Systems (CPSS), a tensor-based multiple clustering (TMC) was proposed. However, as the scale of data continues to increase, data storage, computing load, and memory overhead will increase exponentially, leading to dimensional disasters and greatly affecting the efficiency of TMC. Therefore, a tensor train-based multiple clustering (TTMC) and its parallel computing method are studied in this paper. First, a tensor train (TT)-based multiple clustering parallel analytic and service framework is present. Then, a TT-based multi-linear attribute combination weight learning algorithm, a selective weighted tensor train distance and the TTMC algorithm are put forward to improve the accuracy and efficiency of TMC. Furthermore, an efficient distributed parallel computing strategy of TTMC is designed by using TT core parallelism. Experimental results demonstrate that TTMC and its parallelization can significantly improve computation efficiency and clustering accuracy while reducing the running memory compared to the original TMC algorithm.