Tensor train based machine learning estimator.
Uses existing machine learning estimators to initialize a tensor train decomposition on a particular feature space discretization. Then this tensor train is further optimized with Riemannian conjugate gradient descent.
This library also implements much functionality related to tensor trains, and Riemannian optimization of tensor trains. Finally there is some functionality for turning decision trees and forests into CP/tensor trains.
This software is the companion to the preprint arXiv:2203.04352
The ttml python package can be installed using pip by running
pip install ttml
or by cloning this repository and running the following command in the root directory of this project:
git clone [email protected]:RikVoorhaar/ttml.git
pip install .
If you want to reproduce the experiments discussed in our paper, then first clone this repository. Then run
the script datasets/download_datasets.py to download all relevant datasets from the UCI Machine Learning
Repository. Then all figures and results can be reproduced by the
scripts in the notebooks folder. To install all the dependencies for the scripts and tests, you can use the
conda environment defined in environment.yml.
The documentation for this project lives on ttml.readthedocs.io.
For a more informal explanation of how the machine learning estimator works, see this blog post.
All code for this library has been written by Rik Voorhaar, in a joint project with Bart Vandereycken. This work was supported by the Swiss National Science Foundation under research project 192363.
This software is free to use and edit. When using this software for academic purposes, please cite the following preprint:
@article{
title = {TTML: Tensor Trains for general supervised machine learning},
journal = {arXiv:2203.04352},
author = {Vandereycken, Bart and Voorhaar, Rik},
year = {2022},
}
All figures in the preprint have been produced using version 1.0 of this software.