Search for: All records

In this paper, we use tensor models to analyze the Covid-19 pandemic data. First, we use tensor models, canonical polyadic, and higher-order Tucker decompositions to extract patterns over multiple modes. Second, we implement a tensor completion algorithm using canonical polyadic tensor decomposition to predict spatiotemporal data from multiple spatial sources and to identify Covid-19 hotspots. We apply a regularized iterative tensor completion technique with a practical regularization parameter estimator to predict the spread of Covid-19 cases and to find and identify hotspots. Our method can predict weekly, and quarterly Covid-19 spreads with high accuracy. Third, we analyze Covid-19 data in the US using a novel sampling method for alternating leastsquares. Moreover, we compare the algorithms with standard tensor decompositions concerning their interpretability, visualization, and cost analysis. Finally, we demonstrate the efficacy of the methods by applying the techniques to the New Jersey Covid-19 case tensor data. 

Canonical polyadic (CP) decomposition of a tensor is a basic operation in a lot of applications such as data mining and video foreground/background separation. However, existing algorithms for CP decomposition require users to provide a rank of the target tensor data as part of the input and finding the rank of a tensor is an NP-hard problem. Currently, to perform CP decomposition, users are required to make an informed guess of a proper tensor rank based on the data at hand, and the result may still be suboptimal. In this paper, we propose to conduct CP decomposition and tensor rank approximation together, so that users do not have to provide the proper rank beforehand, and the decomposition algorithm will find the proper rank and return a high-quality result. We formulate an optimization problem with an objective function consisting of a least-squares Tikhonov regularization and a sparse L1-regularization term. We also test its effectiveness over real applications with moving object videos.