The FAµST toolbox provides algorithms and data structures to decompose a given dense matrix into a product of sparse matrices in order to reduce its computational complexity (both for storage and manipulation).
FaµST can be used to:
- speedup / reduce the memory footprint of iterative algorithms commonly used for solving high dimensional linear inverse problems.
- learn dictionaries with an intrinsically efficient implementation
- compute (approximate) fast Fourier transforms on graphs.
A general introduction to FAµST is available in the following paper:
 Le Magoarou L. and Gribonval R., “Flexible multi-layer sparse approximations of matrices and applications”, Journal of Selected Topics in Signal Processing, 2016.
The FAµST toolbox was initially released as a Matlab implementation (versions 1.x) distributed under the terms of the GNU Affero General Public License.
The FAµST toolbox includes a C++ backend compatible with GPU acceleration, and Python / Matlab wrappers.
It is released under a BSD-3-clause license. The Python version is available on PYPI.
Read further news about FAµST.
The development of FAµST was initially funded by the ERC project PLEASE and was further supported by Inria through a Software Development Initiative (ADT REVELATION). Current developments are supported by the ANR Chaire IA AllegroAssai.