L1-Subspace Tracking for Streaming Data
作者：
Highlights：
• We propose an L1-norm principal-component analysis based robust subspace tracking method to capture the intrinsic low-rank structure of streaming data in the presence of outliers.
• The proposed L1-subspace tracking method updates the subspace at each time slot with new sensor datum, utilizing the subspace obtained at the previous time slot, and a small batch of most recent data samples.
• The proposed method has the merits of data outlier suppression through sample weighting and speed acceleration through a warm-start bit-flipping technique, as demonstrated in various application fields.
• The proposed method offers superior subspace estimation accuracy compared to state-of-the-art subspace tracking methods, and is comparable to existing methods in terms of computational complexity and execution time.
摘要
•We propose an L1-norm principal-component analysis based robust subspace tracking method to capture the intrinsic low-rank structure of streaming data in the presence of outliers.•The proposed L1-subspace tracking method updates the subspace at each time slot with new sensor datum, utilizing the subspace obtained at the previous time slot, and a small batch of most recent data samples.•The proposed method has the merits of data outlier suppression through sample weighting and speed acceleration through a warm-start bit-flipping technique, as demonstrated in various application fields.•The proposed method offers superior subspace estimation accuracy compared to state-of-the-art subspace tracking methods, and is comparable to existing methods in terms of computational complexity and execution time.
论文关键词：Dimensionality reduction,Eigenvector decomposition,Internet-of-Things,L1-norm,Outliers,Principal-component analysis,Subspace learning
论文评审过程：Received 9 January 2019, Revised 23 July 2019, Accepted 31 July 2019, Available online 5 August 2019, Version of Record 13 August 2019.
论文官网地址：https://doi.org/10.1016/j.patcog.2019.106992
