How to run decoding

Make sure you set up gsutil first. You should be able to run gsutil ls gs://247-podcast-data without any errors.

Set up

This should be done on a cluster, so ssh first.

git clone git@github.com:hassonlab/247-decoding.git
cd 247-decoding
make download-data

Run

Edit the Makefile to configure your run parameters (subject, pickles, lags, etc.). The options are clearly documented in this file.

Use the following command to run all jobs on a SLURM cluster.

make run-decoding

Use squeue to monitor your jobs.

Results

All results are saved in the results directory. You can use aggregate_results.py to create a CSV summary of all results, or plot.py to plot certain parts of the file. Look at the Makefile's plot target for an example.

Examples

Below are example Makefile configurations. If you set those and run make run-decoding, you'll get the expected output.

625 comprehension only at lag 0

# Choose the subject to run for
SID := 625

# Choose which modes to run for: production, comprehension, or both.
MODES := comp

# Choose the lags to run for.
LAGS := 0

625 production and comprehension at lags -2s to +2s in 1s increments

# Choose the subject to run for
SID := 625

# Choose which modes to run for: production, comprehension, or both.
MODES := prod comp

# Choose the lags to run for.
LAGS = $(shell seq -1024 512 1024)

676 production at lags 0ms 500ms 1s

# Choose the subject to run for
SID := 676

# Choose which modes to run for: production, comprehension, or both.
MODES := prod

# Choose the lags to run for.
LAGS = 0 256 512