Step 0: Environment set up suitable conda environment (assumes conda installed) conda create -n torch python=3.10 -y

conda activate torch

conda install pytorch cpuonly -c pytorch -y

conda install numpy -y

conda install pandas -y

conda install scipy -y

pip install mat73

conda install numba -y

conda install scikit-learn -y

pip install matplotlib

pip install statsmodels

Install git-lfs. Necessary to properly clone files (mac shown here):

brew install git-lfs

git lfs install

git lfs pull

Clone github repository

Step 1: Encoding Sample code to open the template datum: import pandas as pd from utils import load_pickle datum = load_pickle('../template_datum.pkl') df = pd.DataFrame.from_dict(datum)

Replace embeddings in template_datum with your own.

Put embeddings in: /data/pickles/

In Makefile: Change PKL_IDENTIFIER to match your embeddings Change NUM_LAYERS, LAYERS as necessary Change OUT_NAME Change CMD based on your setup. May need your own “submit.sh” file for submitting multiple jobs at once

Then run: make run-layered-sig-encoding

Verification: Sometimes not all encodings run. To check, run: make verify-encoding

Outputs layers that are missing files and the numbers of files that are complete. Re-run those layers and verify again.

Getting Plots: Run: make plot-layered-sig-encoding

Output: Per roi: Inverted u plot Encoding plot Scaled encoding plot Lag layer plot