This class can be used to produce firing rates for any number of simulated motor units according to arbitrary force or kinematic profiles which act as "drive" signals for the simulated motor unit population. The default behavior of the simulator is based on Fuglevand et al., 1993b.
To try the class, you must first install the dependencies. This can be done by configuring an environment with miniconda or micromamba with conda create --name MUsim --file requirements.txt, or micromamba env create -f environment.yaml. You must first install one of these, or use your favorite environment manager.
This class was used to produce the simulations for the main results in the EMUsort paper. To be able to run these simulations directly on your own system, you can download the proper data and generation scripts by switching to either the rat-simulation branch or the monkey-simulation branch with either of the below commands in each section below:
git checkout rat-simulation
You can find example usages of the MUsim class in generate_simulated_dataset_RAT.py, particularly in the sample_MUsim_obj and batch_run_MUsim functions. This generate_simulated_dataset_RAT.py file allows you to generate the exact same spike times with the exact same waveform shapes as in the simulated rat dataset used for all comparisons in the EMUsort paper. Due to non-deterministic operations in the Gaussian noise generation method we used with PyTorch, the continuous*.dat file output will not be exactly the same on your system, so we provide that on Internet Archive (see below). The continuous data, spike times and cluster IDs, and sample numbers files used in the EMUsort paper are all provided there in a ZIP archive under output_files/continuous_dat_files, output_files/spikes_npy_files and output_files/sample_numbers_npy_files, respectively.
When you run generate_simulated_dataset_RAT.py, new files will be generated into the proper subfolders of output_files. Afterwards, you can check for identical matches between the spikes files you generated and the ones provided in the output_files folder using sha256sum (or similar commands). You may check the provided checksum_spikes_npy_RAT.sha256 file to see what the hash should be for the full-sized spikes array. We only provide a checksum file for the large, time-stamped .npy spikes array, but if desired, you could compute the checksum for the spike_clusters.npy and spike_times.npy files we provide at output_files/spikes_npy_files and compare those to the checksums of the corresponding files you've generated.
If you wish, you may also download a file archive containing the full simulation outputs that were used in the EMUsort paper from our Internet Archive listing with the unique identifier: MUsim-rat-simulation (see HERE). The ZIP archive may also be downloaded directly using the following command on Linux systems:
wget https://archive.org/download/MUsim-rat-simulation/rat-simulation_output_files.zip
git checkout monkey-simulation
You can find example usages of the MUsim class in generate_simulated_dataset_MONKEY.py, particularly in the sample_MUsim_obj and batch_run_MUsim functions. This generate_simulated_dataset_MONKEY.py file allows you to generate the exact same spike times with the exact same waveform shapes as in the simulated monkey dataset used for all comparisons in the EMUsort paper. Due to non-deterministic operations in the Gaussian noise generation method used with PyTorch, the continuous*.dat file output will not be exactly the same on your system, so we provide that along with all other outputs on Internet Archive (see below). The spike times and sample numbers files used in the EMUsort paper are also provided in output_files/spikes_npy_files and output_files/sample_numbers_npy_files, respectively.
When you run generate_simulated_dataset_MONKEY.py, new files will be generated into the proper subfolders of output_files. Afterwards, you can check for identical matches between the spikes files you generated and the ones provided in the output_files folder using sha256sum (or similar commands). You may check the provided checksum_spikes_npy_MONKEY.sha256 file to see what the hash should be for the full-sized spikes array. We only provide a checksum file for the large, time-stamped .npy spikes array, but if desired, you could compute the checksum for the spike_clusters.npy and spike_times.npy files we provide at output_files/spikes_npy_files and compare those to the checksums of the corresponding files you've generated.
If you wish, you may also download a file archive containing the full simulation outputs that were used in the EMUsort paper from our Internet Archive listing with the unique identifier: MUsim-monkey-simulation (see HERE). The ZIP archive may also be downloaded directly using the following command on Linux systems:
wget https://archive.org/download/MUsim-monkey-simulation/monkey-simulation_output_files.zip
All data used for computing Tables 1-2 and Figures 6A-B and 7A-B in the EMUsort publication are provided in the EMUsort_paper_supplements folder. The raw outputs used to create these spreadsheets were generated by the plot4() function in compute_ground_truth_metrics.py and are stored inside the EMUsort_paper_supplements/raw_outputs folder.