yuehu9 / Online_Robust_Tensor_Recovery

code for online robust tensor recovery (OLRTR)

Repository from Github https://github.comyuehu9/Online_Robust_Tensor_RecoveryRepository from Github https://github.comyuehu9/Online_Robust_Tensor_Recovery

Online_Robust_Tensor_Recovery (OLRTR)

Streaming data preprocessing via online tensor recovery for large environmental sensor networks Yue Hu Apr. 2021

Overview

This repository contains the source code and tests developed for online robust tensor recovery for urban sensing network data preprocessing. This results are reported in "Streaming data preprocessing via online tensor recovery for large environmental sensor networks" by Y.Hu et. al.

Structures

  • /code/ The source code folder.
    • OLRTR.m is the main algorithm. It solves online tensor robust complepetion under fiber-sparse corruption.
    • solve_proj_21.m and update_L_col.m are helper function for the main algorithm OLRTR.m
    • /PROPACK/ Prerequisit packages, including code for efficient PCA.
    • /tensor_toolbox-master/ is a tensor manipulation package
    • test_numerical_simulate.m is an experiment of OLRTR on numerically simulated tensor data
    • test_NOAA.m is the test code for recoverying the manually corrupted NOAA temperature data.
    • missing_corruptl21.m is the funciton to generate polluted NOAA data.
    • test_AoT.m is the test code for recoverying raw AoT data
    • simulate_tensor.m is the helper function for manually corrupting tensors.
  • /Data/ Contains dataset for testing.
    • /NOAA_12M.mat/ contains the 12 month original NOAA data in Chicago.
    • /aot_12M.mat/ contains the 12 month raw AOT data of 52 sensors.
    • /noaa_chi_12M.mat/ contains the NOAA record of the nearest noaa sensor to the AOT nodes at the same time stamps.

Usage

The code can be run in Matlab. code/test_numerical_simulate.m, test_NOAA.m and test_AoT.m are simulation test, NOAA data test and Chicago AOT case study, respectively.

Contact

  • Author: Yue Hu, Institute for Software Integrated Systems, Vanderbilt University
  • Email: yue.hu (at) vanderbilt.edu

About

code for online robust tensor recovery (OLRTR)

License:MIT License


Languages

Language:MATLAB 59.3%Language:HTML 38.6%Language:XSLT 0.9%Language:Fortran 0.6%Language:C 0.4%Language:CSS 0.2%