Convolutional Dictionary Learning with Grid Refinement for Spike Sorting

Mini-Project for the MVA class ML for Timeseries, based on the study of the article:

Convolutional sparse coding (CSC) and convolutional dictionary learning (CDL) for off-the-grid events. Song, A., Flores, F., and Ba D., Convolutional Dictionary Learning with Grid Refinement, IEEE Transaction on Signal Processing, 2020

Report is available here.

The parts of the code coming from srcdl are highlighted with comments.

Installation

Please create a specific Python 3.11 environment, and then install the cdlgr package: pip install -e .. You will then be able to run the command ssrun from the package.

Note that you may need to set up Kachery Cloud (i.e. link it with your Github account) to be able to download spikeforest datasets, with the command kachery-cloud-init (if automatic configuration fails).

Use

Please follow the notebook cdlgr.ipynb to see how to run the code and view the results.

It has been tested in VS Code.