Reproducibility Instrucion for "Multivariate Low-Rank State-Space Model with SPDE Approach for High-Dimensional Data"
Author: Jacopo Rodeschini, Lorenzo Tedesco
This file documents the code associated with the simulation reported in the article and outputs supporting the computational findings and describes how to reproduce simulation results.
This paper proposes a novel low-rank approximation to the multivariate State-Space Model. The Stochastic Partial Differential Equation (SPDE) approach is applied component-wise to the independent-in-time Matérn Gaussian innovation term in the latent equation, assuming component independence. This results in a sparse representation of the latent process on a finite element mesh, allowing for scalable inference through sparse matrix operations. Dependencies among observed components are introduced through a matrix of weights applied to the latent process. Model parameters are estimated using the Expectation-Maximisation algorithm, which features closed-form updates for most parameters and efficient numerical routines for the remaining parameters. We prove theoretical results regarding the accuracy and convergence of the SPDE-based approximation under fixed-domain asymptotics. Simulation studies show our theoretical results. We include an empirical application on air quality to demonstrate the practical usefulness of the proposed model, which maintains computational efficiency in high-dimensional settings. In this application, we reduce computation time by about 93%, with only a 15% increase in the validation error.
-
LRSSM_simulations.py: Run the simulations to confirm theoretical findings. Results are saved in theoutput/folder. -
LRSSM_case_study.py: Run the case study application. Results are saved in theoutput/folder. -
LRSSM_comparison.py: Run the LRSSM for the comparison. Results are saved in theoutput/folder. -
utils/: Module containing utility functions. -
output/: Stores simulation results. -
shapefile/: Folder with the Italian shapefile required for the case study. -
environment.txt: List of Python packages required to create the conda environment. -
COMPARISON_INLA.R: Run the SSM with INLA for the comparison. -
COMPARISON_INLADIFF.R: Run the diffuseve model with INLA for the comparison.
-
Download the repository as a ZIP file named
Low_Rank_State_Space_Model.zip, and extract it into a folder namedLow_Rank_State_Space_Model. -
Navigate into the folder:
cd Low_Rank_State_Space_Model- Create the Conda environment (named
dev) from the list of required packages:
conda env create --file=environment.yml --name dev- Activate the environment:
conda activate dev- Activate the environment:
conda activate dev- Run the
LRSSM_simulations.pyscript:
python LRSSM_simulations.py-
Extract the dataset (
.csv) from empirical_application.zip under the ./output folder. It comes from the Zenodo repository: GRINS_AQCLIM_points_Italy.rda. -
Activate the environment:
conda activate dev- Run the case study using the Python script:
python LRSSM_case_study.py - Activate the environment:
conda activate devpython LRSSM_comparison.py- Run the
COMPARISON_INLA.Rscript:
Rscript COMPARISON_INLA.R
- Run the
COMPARISON_INLADIFF.Rscript
Rscript COMPARISON_INLADIFF.R
The computation time reported in the paper was recorded using Python 3.10.15 on a machine equipped with two Intel(R) Xeon(R) Platinum 8460Y+ CPUs and 1007 GB RAM.