MPSTime.jl

A Julia package for time-series machine learning (ML) using Matrix-Product States (MPS) built on the ITensors.jl framework [1, 2].

Overview

MPSTime is a Julia package for learning the joint probability distribution of time series directly from data using matrix product state (MPS) methods inspired by quantum many-body physics. It provides a unified formalism for:

• Time-series classification (inferring the class of unseen time-series).
• Univariate time-series imputation (inferring missing points within time-series instances) across fixed-length time series.
• Synthetic data generation (coming soon).

Installation

This is not yet a registered Julia package, but it will be soon (TM)! In the meantime, you can install it directly from our GitHub repository:

julia> ]
pkg> add https://github.com/joshuabmoore/MPSTime.jl.git 

Usage

See the sidebars for basic usage examples. We're continually adding more features and documentation as we go.

Citation

If you use MPSTime in your work, please read and cite the arXiv preprint:

@misc{MPSTime2024,
      title={Using matrix-product states for time-series machine learning}, 
      author={Joshua B. Moore and Hugo P. Stackhouse and Ben D. Fulcher and Sahand Mahmoodian},
      year={2024},
      eprint={2412.15826},
      archivePrefix={arXiv},
      primaryClass={stat.ML},
      url={https://arxiv.org/abs/2412.15826}, 
}