fastmat

Description

Scientific computing requires handling large composed or structured matrices. Fastmat is a framework for handling large composed or structured matrices. It allows expressing and using them in a mathematically intuitive way while storing and handling them internally in an efficient way. This approach allows huge savings in computational time and memory requirements compared to using dense matrix representations.

Dependencies

• Python 2.7, Python >=3.5
• Numpy >= 1.16.3
• Scipy >= 1.0
• Cython >= 0.29
• soft dependencies:
  • matplotlib: for demos and tools that make use of plotting functions

Distribution

Binary wheels are provided for Python >=3.5 for linux, windows and mac, as well as for x86 and ARM architectures.

For all systems, for which no wheels are provided, you may still install fastmat from the soruce distribution.

Authors & Contact Information

• Sebastian Semper | sebastian.semper@tu-ilmenau.de Technische Universität Ilmenau, Institute for Mathematics, EMS Group
• Christoph Wagner | christoph.wagner@tu-ilmenau.de Technische Universität Ilmenau, Institute for Information Technology, EMS Group
• https://www.tu-ilmenau.de/it-ems/

Citation / Acknowledgements

If you use fastmat, or parts of it, for commercial purposes you are required to acknowledge the use of fastmat visibly to all users of your work and put a reference to the project and the EMS Group at TU Ilmenau.

If you use fastmat for your scientific work you are required to mention the EMS Group at TU Ilmenau and cite the following publication affiliated with the project:

Christoph W. Wagner and Sebastian Semper and Jan Kirchhof, fastmat: Efficient linear transforms in Python, SoftwareX, 2022, https://doi.org/10.1016/j.softx.2022.101013

    @article{Wagner_2022,
        doi = {10.1016/j.softx.2022.101013},
        url = {https://doi.org/10.1016%2Fj.softx.2022.101013},
        year = {2022},
        month = {jun},
        publisher = {Elsevier {BV}},
        volume = {18},
        pages = {101013},
        author = {Christoph W. Wagner and Sebastian Semper and Jan Kirchhof},
        title = {fastmat: Efficient linear transforms in Python},
        journal = {{SoftwareX}}
    } 

• https://www.tu-ilmenau.de/it-ems/

Installation

fastmat currently supports Linux, Windows and Mac OS. Lately it also has been seen on ARM cores coming in a Xilinx ZYNQ FPGA SoC shell. We encourage you to go ahead trying other platforms as the aforementioned as well and are very happy if you share your experience with us, allowing us to keep the list updated.

Installing with pip:

fastmat is included in the Python Package Index (PyPI) and can be installed from the commandline by running one easy and straightforward command: pip install fastmat

When installing with pip all dependencies of the package will be installed along. With release 0.1.1 python wheels will be offered for many versions greatly improving installation time and effort.

Bulding from source

Building binaries has been developed and tested for the use

Manually installing from source

• download the source distribution from our github repository: https://github.com/EMS-TU-Ilmenau/fastmat/archive/stable.zip
• unpack its contents and navigate to the project root directory
• run pip install . to install fastmat on your computer
• you may also install fastmat without pip, using the offered makefile:
  • type make install to install fastmat on your computer
  • If you intend to install the package locally for your user type make install MODE=--user instead
  • You may add a version specifier for all make targets that directly or indirectly invoke Python: make install PYTHON=python2 make compile PYTHON=python3
  • If you only would like to compile the package to use it from this local directory without installing it, type make compile
  • An uninstallation of a previously run make installis possible, provided the installation log file setup.files has been preserved Invoking make uninstall without a local setup.files causes another installation for generating the setup file log prior to uninstalling
• NOTE: Windows users If you intent on building fastmat from source on a windows platform, make sure you have installed a c compiler environment and make interpreter. One way to accomplish this is to install these tools for Python 2.7 (you may also chose different ones, of course):
  • Intel Distribution for Python 2.7
  • Microsoft Visual C++ Compiler 9.0 for Python 2.7
  • GNU make for Windows 3.81 or newer
  • depending on your system: The relevant header files

Demos

Feel free to have a look at the demos in the demo/ directory of the source distribution. Please make sure to have fastmat already installed when running these.

Please note that the edgeDetect demo requires the Python Imaging Library (PIL) installed and the SAFT demos do compile a cython-core of a user defined matrix class beforehand thus having a delaying the first time they're executed.

Documentation / HELP !

Please have a look at the documentation, which is included in the source distribution at github or may be built locally on your machine by running make doc

If you experience any trouble please do not hesitate to contact us or to open an issue on our github projectpage: https://github.com/EMS-TU-Ilmenau/fastmat

FAQ

Please check out our project documentation at readthedocs.

Windows: Installation fails with various "file not found" errors

Often, this is caused by missing header files. Unfortunately windows ships without a c-compiler and the header files necessary to compile native binary code. If you use the Intel Distribution for Python this can be resolved by installing the Visual Studio Build tools with the version as recommended by the version of the Intel Distribution for Python that you are using.

Issue not resolved yet?

Please contact us or leave your bug report in the issue section. Thank You!