Abstract:As a regression technique in spatial statistics, the spatiotemporally varying coefficient model (STVC) is an important tool for discovering nonstationary and interpretable response-covariate associations over both space and time. However, it is difficult to apply STVC for large-scale spatiotemporal analyses due to its high computational cost. To address this challenge, we summarize the spatiotemporally varying coefficients using a third-order tensor structure and propose to reformulate the spatiotemporally varying coefficient model as a special low-rank tensor regression problem. The low-rank decomposition can effectively model the global patterns of large data sets with a substantially reduced number of parameters. To further incorporate the local spatiotemporal dependencies, we use Gaussian process (GP) priors on the spatial and temporal factor matrices. We refer to the overall framework as Bayesian Kernelized Tensor Regression (BKTR), and kernelized tensor factorization can be considered a new and scalable approach to modeling multivariate spatiotemporal processes with a low-rank covariance structure. For model inference, we develop an efficient Markov chain Monte Carlo (MCMC) algorithm, which uses Gibbs sampling to update factor matrices and slice sampling to update kernel hyperparameters. We conduct extensive experiments on both synthetic and real-world data sets, and our results confirm the superior performance and efficiency of BKTR for model estimation and parameter inference.

What problem does this paper attempt to address?

### The Problem Addressed by the Paper This paper aims to address the high computational cost of the Spatio-Temporal Varying Coefficient (STVC) model in large-scale spatio-temporal data analysis. Specifically, traditional STVC methods face the challenge of high computational costs when dealing with large datasets, which limits their widespread use in practical applications. To overcome this challenge, the authors propose a new framework based on Bayesian Kernel Tensor Regression (BKTR). ### Main Contributions 1. **Low-Rank Tensor Decomposition**: By representing the spatio-temporal varying coefficients as a third-order tensor and using low-rank tensor decomposition to model global patterns, the number of parameters is significantly reduced. 2. **Local Spatio-Temporal Dependency**: By using Gaussian Process (GP) priors on the spatial and temporal factor matrices, local spatio-temporal dependencies are further incorporated. 3. **Efficient Inference Algorithm**: An efficient Markov Chain Monte Carlo (MCMC) algorithm is developed to update the factor matrices and kernel hyperparameters. 4. **Scalability and Robustness**: Through experimental validation, BKTR demonstrates superior performance and efficiency in model estimation and parameter inference, capable of handling large-scale spatio-temporal data and multidimensional relationships. ### Method Overview - **Model Definition**: The spatio-temporal varying coefficient model is reformulated as a low-rank tensor regression problem, using low-rank decomposition to model global patterns. - **Bayesian Framework**: Gaussian Process priors are used on the spatial and temporal factor matrices, assuming that the columns of the factor matrix \( W \) follow a zero-mean Gaussian distribution. - **MCMC Algorithm**: Gibbs sampling is used to update the factor matrices, and slice sampling is used to update the kernel hyperparameters. - **Computational Complexity**: By using low-rank tensor decomposition, the computational complexity is reduced from \( O(M^3N^3P^3) \) to \( O(R^3(M^3 + N^3 + P^3)) \), where \( R \) is the rank of the tensor. ### Experimental Results The authors conducted extensive experiments on synthetic and real datasets, showing that BKTR exhibits superior performance and efficiency in model estimation and parameter inference, particularly excelling in handling large-scale spatio-temporal data. ### Conclusion The BKTR framework proposed in this paper provides an efficient and scalable solution for large-scale spatio-temporal data analysis, significantly reducing the computational cost of traditional STVC methods, making them more feasible for practical applications.

Scalable Spatiotemporally Varying Coefficient Modelling with Bayesian Kernelized Tensor Regression

Spatiotemporal Covariance Neural Networks

Large Bayesian Tensor VARs with Stochastic Volatility

A Spatiotemporal Regression-Kriging Model for Space-Time Interpolation: a Case Study of Chlorophyll-a Prediction in the Coastal Areas of Zhejiang, China.

Variational Bayesian Inference for Tensor Robust Principal Component Analysis

Nonstationary Spatial Process Models with Spatially Varying Covariance Kernels

Scalable Bayesian Tensor Ring Factorization for Multiway Data Analysis

Multivariate Spatial-temporal Prediction on Latent Low-dimensional Functional Structure with Non-stationarity

Bayesian Nonstationary Spatial Modeling for Very Large Datasets

TVGCN: Time-varying graph convolutional networks for multivariate and multifeature spatiotemporal series prediction

Bayesian Kernelized Tensor Factorization as Surrogate for Bayesian Optimization

Bayesian tensor-on-tensor regression with efficient computation

Fast and Flexible Bayesian Inference in Time-varying Parameter Regression Models

Spatio-Temporal Dual Kriging with Adaptive Coefficient Drift Function

Is Single Enough? A Joint Spatiotemporal Feature Learning Framework for Multivariate Time Series Prediction

Modeling Multivariate Spatial-Temporal Data with Latent Low-Dimensional Dynamics

Multivariate Spatiotemporal Models with Low Rank Coefficient Matrix

Spatially Clustered Varying Coefficient Model

Sparse spatially clustered coefficient model via adaptive regularization

Bayesian Temporal Factorization for Multidimensional Time Series Prediction

Bayesian spatially varying coefficient models in the spBayes R package