Overview

A Python package for the fast calculation of Multi-unit Van Rossum neural spike train metrics, with the kernel-based algorithm described in Houghton and Kreuz, On the efficient calculation of Van Rossum distances (Network: Computation in Neural Systems, 2012, 23, 48-58). This package started out as a Python wrapping of the original C++ implementation given by the authors of the paper, and evolved from there with bugfixes and improvements.

Documentation

Full documentation is hosted at http://pymuvr.readthedocs.org/.

Requirements

• Python 2.7 or 3.x.

• NumPy>=1.7.

• C++ development tools and Standard Library (package build-essential on Debian).

• Python development tools (package python-dev on Debian).

Installation

To install the latest release, run:

pip install pymuvr

If you prefer installing from git, use:

git clone https://github.com/epiasini/pymuvr.git
cd pymuvr
python setup.py install

Usage

The module exposes two functions:

pymuvr.distance_matrix(observations1, observations2, cos, tau)

pymuvr.square_distance_matrix(observations, cos, tau)

distance_matrix calculates the ‘bipartite’ (rectangular) dissimilarity matrix between the multi-unit trains in observations1 and those in observations2.

square_distance_matrix calculates the ‘all-to-all’ dissimilarity matrix between each pair of trains in parallel_trains. It’s an optimised form of distance_matrix(observations, observations, cos, tau).

They both return their results as a 2D numpy array.

The observations arguments must be thrice-nested lists of spiketimes, in such a way that observations[i][j][k] represents the time of the kth spike of the jth cell of the ith observation. cos and tau are the usual parameters for the multiunit Van Rossum metric.

See examples/benchmark_versus_spykeutils.py for an example of usage comparing the performance of pymuvr with the pure Python implementation of the multiunit Van Rossum distance in spykeutils.

License

This package is licensed under version 3 of the GPL or any later version. See COPYING for details.

Getting the source

Source code for pymuvr is hosted at https://github.com/epiasini/pymuvr.