Skip to content

This repository contains MATLAB files for the implementation of work proposed in the paper Efficient Structure-preserving Support Tensor Train Machine.

License

Notifications You must be signed in to change notification settings

mpimd-csc/Structure-preserving_STTM

Repository files navigation

Efficient Structure-preserving Support Tensor Train Machine

This repository contains MATLAB files for the implementation of work proposed in the paper Efficient Structure-preserving Support Tensor Train Machine.

Intro

The key novelty of our research is a stable and well explained Support Vector Machine (SVM) model for low-rank tensor input data that manifests much higher classification accuracy and banchmarked compared to other state-of-the-art methods. Our paper presents a general SVM framework using the Tensor-Train decomposition along with the explanation, validation and importance of each stage of the proposed algorithm with a graphical illustration.

Dataset

Folder - datasets

We have taken two different types of datasets. One medical data (resting-state fMRI) and another Hyperspectral Images.

Medical resting-state fMRI Data

ADNI_first (Alzheimer disease) and ADHD (Attention Deficit Hyperactivity Disorder)

Hyperspectral Images

Indian Pines and Salinas

Setup

Libraries:

  1. Tensor-Train Toolbox by Ivan Oseledets and Sergey Dolgov
  2. LIBSVM by Chih-Chung Chang and Chih-Jen Lin

Functions and Results

Each folder presents results for each step of algorithm, presented in paper.

Comparision of our method to state-of-the-art -> run the file named Mainfile_results.m in the 5th folder.

Cite As

If you use our work and codes for the further research then please cite the paper [Efficient_STTM].

BibTeX
@article{JMLR:v24:20-1310,
  author  = {Kirandeep Kour and Sergey Dolgov and Martin Stoll and Peter Benner},
  title   = {Efficient Structure-preserving Support Tensor Train Machine},
  journal = {Journal of Machine Learning Research},
  year    = {2023},
  volume  = {24},
  number  = {4},
  pages   = {1--22},
  url     = {http://jmlr.org/papers/v24/20-1310.html}
}
}

If you have any query/suggestion, kindly write to Kirandeep Kour at kour@mpi-magdeburg.mpg.de.

About

This repository contains MATLAB files for the implementation of work proposed in the paper Efficient Structure-preserving Support Tensor Train Machine.

Topics

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published