This repository contains code for the paper Distributed-Memory Randomized Algorithms for Sparse CP Decomposition, to appear at SPAA 2024.

You can test two randomized algorithms, CP-ARLS-LEV and STS-CP, that can decompose billion-scale sparse tensors on a cluster of CPUs.

Clone the repo and cd into it.

[zsh]> git clone https://github.com/vbharadwaj-bk/fast_tensor_leverage.git
[zsh]> cd fast_tensor_leverage

Install Python dependencies with the following command:

[zsh]> pip install -r requirements.txt

We rely on the Pybind11 and cppimport packages. We use the HDF5 format to store sparse tensors, so you need the h5py package if you want to perform sparse tensor decomposition.

Within the repository, run the following command:

[zsh]> python configure.py