anabatsh / TT_pro

TT_pro

Description

Method PROTES (PRobability Optimizer with TEnsor Sampling) for optimization of the multidimensional arrays and discretized multivariable functions based on the tensor train (TT) format.

---

Attention! This is a repository for code development and testing. Please use the stable version of our optimizer from the protes repository (it can be installed as pip install protes).

---

Installation

1. Install python (version 3.8; you may use anaconda package manager);

2. Create a virtual environment:

   conda create --name tt_pro python=3.8 -y

3. Activate the environment:

   conda activate tt_pro

4. Install dependencies:

   pip install numpy scipy teneva==0.12.8 ttopt==0.5.0 jax optax equinox qubogen gekko nevergrad torch

5. Clean temporary dir after runs:

   find /tmp -type d -maxdepth 1 -iname "*model*" -exec rm -fr {} \;

6. Delete virtual environment at the end of the work (optional):

   conda activate && conda remove --name tt_pro --all -y

Usage

Please, see our colab notebook with various examples.

Authors

• Anastasia Batsheva
• Ivan Oseledets
• Andrei Chertkov
• Gleb Ryzhakov

Citation

If you find our approach and/or code useful in your research, please consider citing:

@article{batsheva2023protes,
    author    = {Batsheva, Anastasia and Chertkov, Andrei  and Ryzhakov, Gleb and Oseledets, Ivan},
    year      = {2023},
    title     = {PROTES: Probabilistic Optimization with Tensor Sampling},
    journal   = {arXiv preprint arXiv:2301.12162},
    url       = {https://arxiv.org/pdf/2301.12162.pdf}
}