jieli-matrix/HOEDMD

HOEDMD Code

Core Algorithms

Implementation of core algorithms (HOEDMD-STLS, HOEDMD-NONSVD and HOEDMD-KER) has been put under algs folder. We also implement HODMD algorithm as hodmd_v3.m.

(base) ➜ HOEDMD tree algs                       
algs
├── hodmd_v3.m
├── hoedmd_ker.m
├── hoedmd_stls.m
└── kermthd
    ├── ker_vec.m
    ├── kerprod.m
    └── opt_solve.m

Multilinear PageRank

The following code indicates how to reproduce the results in the multilinear pagerank example and visualize it.

addpath(genpath('./algs'))
addpath('./utils')
addpath(genpath('./multi-pr'))
cd './multi-pr'
% generate hoedmd results
clc;clear;
% set N = 50; M = 2; in the 1st line in main.m and main_plot.m
main
main_plot

Time series classification

The following code indicates how to reproduce the results in the time series classification example and analyze it. Data-Preprocess Notes: Please navigate to BasicMotions to download the dataset and preprocess each sample into mat format. Edit files' location information in the config file clus_cfgs.mat. To apply DMD on time series

addpath(genpath('./algs'))
addpath('./utils')
addpath(genpath('./ucr'))
cd './ucr/cluster'
bm_ker_vap_p

The DMD results will generate under the data/hoedmd/BM, then you can run the cluster with elbow.

run_cluster_ker

You can analyze the results by visualize elbow_table_ker. Each column represents Accuracy (ACC), Sensitivity (SEN), and Specificity (SPE); Each row represents p performance from 10 to 90 under the S = 10.

fMRI data

Data-Preprocess Notes:Data used in this section were provided by the Human Connectome Project, WU-Minn Consortium. We randomly select one subject and apply DMD variants to his/her fMRI signal, which is saved in fmri_table. We replace the file with a random matrix for the consideration of HCP data license, so please replace the file with the fMRI data downloaded from HCP when reproducing the example.

To generate DMD-variants results, you should install PyDMD as PyDMD Project Indicates.

python main_dmd.py

It will generate DMD_dict.mat in the folder.

addpath(genpath('./algs'))
addpath('./utils')
addpath(genpath('./fmridmd'))
cd './fmridmd'
gd_hoedmd

You can view the visulization by boxchart

boxchart(abs(real(log(eigs_table_S1))))
boxchart(abs(real(log(eigs_table_S2))))