Spike Pattern Optimal Transport

The Python 3 module spot implements the Spike Pattern Optimal Transport Dissimilarity described in Grossberger, L., Battaglia, F. and Vinck, M. (2018). Unsupervised clustering of temporal patterns in high-dimensional neuronal ensembles using a novel dissimilarity measure. PLoS Comput. Biol.

Setup

The dependencies can be installed by running ./env_setup.sh <ENV_NAME> with the optional argument specifying the target environment (which must be source-able).

To setup the module, run python setup.py install.

A jupyter notebook is available in notebooks/, along with a demo dataset, showing an example workflow for the SPOTDisClust method.

Linux (Anaconda)

1. conda create -n spot python=3
2. ./env_setup.sh spot
3. source activate spot
4. python setup.py install

Windows (Anaconda)

1. conda create -n spot python=3
2. conda activate spot
3. conda install python=3.6.5
4. conda install -c conda-forge hdbscan=0.8.13=py36_0
5. conda install numba
6. conda install ipykernel
7. conda install matplotlib
8. python setup.py install

Loading data from a MATLAB workspace

For your convenience, the script scripts/from_matlab.py -i example_data/matlab_workspace_example.mat -o example_data/matlab_ converts spike data from a MATLAB (.mat) workspace file, specified by -i, to the format required by SPOTDis and saves the resulting files with the file name prefix, specified by -o. For this script to work, make sure that it contains the following variables (or follow the example_data/generate_matlab_workspace_example.m, which generated the MATLAB example):

• neuron_spike_times cell array with one entry per neuron containing a vector of spike times
• trial_start_times vector containing start times for all trials (inclusive)
• trial_end_times vector containing end times for all trials (exclusive)