pei-jie/ttpy

Python implementation of the TT-Toolbox

ttpy

Python implementation of the TT-Toolbox. It contains several important packages for working with the Tensor Train (TT) format in Python. It is able to do interpolation, solve linear systems, eigenproblems, solve dynamical problems. Several computational routines are done in Fortran (which can be used separatedly), and are wrapped with the f2py tool.

Main contributors

• Ivan Oseledets (version 0.1, build system, CROSS, KSL and EIGB modules)
• Tigran Saluev (made the code usable; implemented multifucrs and TT-GMRES algorithms and many more; a lot of bugfixes; initial documentation)
• Dmitry Savostyanov ( tt-fort submodule code, AMEN module)
• Sergey Dolgov ( AMEN module)

Installation

##From a binstar repository If you use 64 bit linux, you can install the module from the binstar repository:

sudo apt-get install libgfortran3
conda install -c https://conda.binstar.org/bihaqo ttpy

##Downloading the code This installation works with git submodules, so you should be sure, that you got them all right. The best way is to work with this repository is to use git with a version >= 1.6.5. Then, to clone the repository, you can simply run

git clone --recursive git://github.com/oseledets/ttpy.git

To update to the latest version, run

git pull
git submodule update --init --recursive *
python setup.py install

This command will update the submodules as well.

##Prerequisites It is highly recommended that you use either

• Anaconda Python distribution which has MKL in built in for the academics Anaconda Python is the version for which the development of ttpy is done

• Enthought Python distribution -- should work as well with the non-free version, but not tested

##Installing the package The installation of the package is done via setup.py scripts.

The installation script is a very newbie one, but it seems to work. Go to the tt directory and run

python setup.py install

This builds the package tt with submodules amr, eigb, kls. For some of the examples the quadgauss package is required, so go to quadgauss directory and run setup.py there

BLAS

Almost all of the packages depend on the BLAS/LAPACK libraries, but the code is not explicitly linked against them, so if you do not have the BLAS/LAPACK in the global namespace of your Python interpreter, you will encounter "undefined symbols" error during the import of the tt package. If you have Anaconda or EPD with MKL installed, it should work ``as it is''. The initialization file of ttpy tries to dynamically load runtime MKL or lapack libraries (should work on Linux if those libraries are in LD_LIBRARY_PATH).

What those packages do

They have the following functionality

• tt : The main package, with tt.tensor and tt.matrix classes, basic arithmetic, norms, scalar products, rounding full -> tt and tt -> full conversion routines, and so on

• tt.amen : AMEN solver for linear systems (Python wrapper for Fortran code written by S. V. Dolgov and D. V. Savostyanov) it can be also used for fast matrix-by-vector products.

• tt.eigb : Block eigenvalue solver in the TT-format (subroutine tt.eigb.eigb)

• tt.ksl : Solution of the linear dynamic problems in the TT-format, using the projector-splitting KSL scheme. A Python wrapper for a Fortran code (I. V. Oseledets)

• tt.cross : Has a working implementation of the black-box cross method. For now, please use the rect_cross function.

Documentation and examples

The package provides Sphinx-generated documentation. To build HTML version, just do

cd tt/doc
make html

A few examples are available right now under examples/ directory

For any questions, please feel free to contact me by email or create an issue on Github.