Tool for estimation of statistical characteristics of multivariate random functions
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.settings
docs
etsproxy
examples
spirrid
unittests
.gitignore
.project
MANIFEST.in
licence.txt
readme.rst
rv_view.py
setup.py

readme.rst

SPIRRID

The SPIRRID package is the part of the project Simvisage (https://github.com/simvisage/spirrid). Use the zip button to download the source code for testing and see the Examples section below. Examples are accompanying and extending the studies provided in the prepared paper Using Python for scientific computing: efficient and flexible evaluation of the statistical characteristics of functions with mutivariate random inputs, R. Chudoba, V. Sadilek, R. Rypl, and M. Vorechovsky, prepared for submission in CPC.

Installation instructions

Prerequisites

Windows, Linux and Mac

The Enthought Python Distribution (EPD) is available in system independent form at

http://www.enthought.com/products/epd.php

On windows, there is some Cython problem (pyximport). At this time, we don't have any simple solution. Due to this fact, it was removed from benchmark tests.

There is a commercial and academic version of EPD containing all required packages. Free distribution of EPDFree does not contain all packages (mayavi, numexpr, cython). These packages must be installed manually.

UBUNTU 10.04

Enthought Tool Suite (ETS 3) and additional utilities can be installed using the standard package manager as:

$ sudo apt-get install ipython python-traitsui python-scipy \
        python-matplotlib mayavi2 cython

In order to make the comparison between numexpr and numpy efficiency in the script examples.numexpr_test.numexpr_test the python-numpexpr package must be installed separately as it is not included in ubuntu 10.04 distribution.

UBUNTU 11.04

Enthought Tool Suite (ETS 4) and additional utilities are installed as:

$ sudo apt-get install mayavi2 python-matplotlib \
    cython python-numexpr python-scipy

Other distributions

For other systems and distributions use the description provided at and references there:

http://blog.enthought.com/open-source/ets-4-0-released/

Download and install

Download and unzip the package from https://github.com/simvisage/spirrid.

Examples

To test spirrid package on prepared examples change to the top directory of the spirrid package run:

$ python examples/demo.py

you get the user interface to run one of examples described in the last section this document.

In order to start the individual examples the top level directory of spirrid package must be included in the PYTHONPATH environment variable or in the sys.path variable of the executed script.

Package structure

spirrid/

This folder contains tools for random variable domain sampling, code generation and numerical multidimensional statistical integration.

spirrid/pdistrib (library of statistical distributions)

The package provides a traited wrapper for the scipy distributions.

spirrid/docs

Package generating the documentation from the source code and from the demonstration examples. See the docs/readme.rst file for further details on generating the documentation. The current documented is available in html format online at http://mordred.imb.rwth-aachen.de/docs/spirrid

spirrid/etsproxy

Subsidiary package needed to support both ETS 3 and ETS 4 with changed import paths.

spirrid/util

This folder contains customized (extended) source code (enthought, scipy, numpy) needed for running spirrid.

examples/

The directories "fiber_*" provides the performance studies of the spirrid integration tool. There are three types of response functions tested:

  • fiber_tt_2p/: fiber tensile test with 2 parameters one strong discontinuity)
  • fiber_tt_5p/: fiber tensile test with 5 parameters one strong and one weak discontinuity
  • fiber_po_8p/: fiber pullout test with 7 parameters one strong, one weak discontinuity and nonlinear range within the response

There are two tests:

  • masked_arrays/: testing of speeding up of evaluation of general function using numpy.ma.array
  • numexpr/: testing of speeding up of evaluation of fiber_tt_5p fiber tensile test with 5 parameters one strong and one weak discontinuity) using numexpr

And:

  • script.py: simple python script, demonstrating several possible ways how to implement the estimation of mean value of a multi-variate random function. The script uses a two-parametric function with a discontinuity (stress-strain response of a fiber loaded in tension). Both parameters of the function are considered randomly distributed.

    The script shows a figure containing two diagrams: The left diagram displays the obtained mean response of the random process for four implemented sampling techniques indluding regular grids and Monte-Carlo types of sampling. The right diagram visualizes the coverage of the random domain with two random variables for the four applied sampling techniques.

    More detailed issues concerning the efficiency of the covered sampling and implementation techniques are described in paper

    CHUDOBA, R.; SADÍLEK, V.; RYPL, R.; VOŘECHOVSKÝ, M. Using Python for scientific computing: Efficient and flexible evaluation of the statistical characteristics of functions with multivariate random inputs. COMPUTER PHYSICS COMMUNICATIONS, 2012, vol. 184, n. 2, pages 414-427. ISSN: 0010- 4655, 10.1016/j.cpc.2012.08.021.