Learning representations of multivariate time series with missing data. (December 2019)

• Record Type: Journal Article
• Authors: Bianchi, Filippo Maria; Livi, Lorenzo; Mikalsen, Karl Øyvind; Kampffmeyer, Michael; Jenssen, Robert
• Is Part Of: Pattern recognition. Volume 96(2019:Dec.)
• Publication Date: 2019-12
• Subjects: Representation learning -- Multivariate time series -- Autoencoders -- Recurrent neural networks -- Kernel methods; Pattern perception -- Periodicals; Perception des structures -- Périodiques; Patroonherkenning; 006.4
• DOI: 10.1016/j.patcog.2019.106973
• Languages: English
• ISSNs: 0031-3203

Abstract

Highlights: We design a recurrent autoencoder architecture to compress multivariate time series with missing data. An additional regularization term aligns the learned representations with a prior kernel, which accounts for missing data. Even with many missing data, time series belonging to different classes become well separated in the induced latent space. We exploit the proposed architecture to design methods for anomaly detection and for imputing missing data. We perform an analysis to investigate which kind of time series can be effectively encoded using recurrent layers. Abstract: Learning compressed representations of multivariate time series (MTS) facilitates data analysis in the presence of noise and redundant information, and for a large number of variates and time steps. However, classical dimensionality reduction approaches are designed for vectorial data and cannot deal explicitly with missing values. In this work, we propose a novel autoencoder architecture based on recurrent neural networks to generate compressed representations of MTS. The proposed model can process inputs characterized by variable lengths and it is specifically designed to handle missing data. Our autoencoder learns fixed-length vectorial representations, whose pairwise similarities are aligned to a kernel function that operates in input space and that handles missing values. This allows to learn good representations, even in the presence of a significant amount of missing data. To show the effectiveness of the proposed approach, we evaluate the quality of the learned representations in several classification tasks, including those involving medical data, and we compare to other methods for dimensionality reduction. Successively, we design two frameworks based on the proposed architecture: one for imputing missing data and another for one-class classification. Finally, we analyze under what circumstances an autoencoder with recurrent layers can learn better compressed representations of MTS than feed-forward architectures.