This GitHub repository contains the python implementation of the Tensor-Train Incremental Core Expansion (TT-ICE) and its heuristically enhanced version TT-ICE*.
The TT-ICE algorithm is implemented as a part of the DaMAT library. More information on DaMAT library can be found in the documentation page
DaMAT is a python library that provides support for tensor operations for Tensor-Train format. You can find both one-shot algorithms (such as TT-SVD) as well as incremental algorithms (such as TT-ICE, TT-FOA and ITTD). Although multiple options were provided for incremental computation of Tensor-Train format, the main aim of this library is to provide a python implementation for TT-ICE.
In addition to the main library/codebase for TT-ICE, this repository also contains scripts that were used for the numerical experiments of the manuscript An Incremental Tensor Train Decomposition Algorithm under the examples folder.
As a part of the reproducibility badge requirements from SIAM SISC, we have submitted the necessary data to reproduce the experiments in Section 4 of the manuscript. You can temporarily access the files through Google Drive until that data is released.
Once you clone this repository, you need to setup the DaMAT package using the following prompt.
python3 setup.py install --user
Please carefully review the associated readme file for the experiments.