eag / esn

command line tool designed to predict epoched EEG data

esn

Purpose

to predict epoched EEG data some "k" steps ahead

Quick Start

Input

• input should be stored as 64 bit floats in a binary file
• first value in the binary file stores the number of timepoints in the epoch
• second value stores the number of epochs in the file
• remaining values store the epoched data for a single channel or multiple channels
• values are generally normalized to lie between 0 and 1

Output

• first value in the binary file stores the number of timepoints in the epoch
• second stores the number of epochs in the file
• third value stores the number of prediction steps
• remaining values store the prediction k steps ahead

Execution

• see https://github.com/eag/esn/blob/main/bash/run_esn for an example

Example output

• 20 step prediction (40 ms) of eeg activity in delta band (obtained using MATLAB's "wavedec" function): https://github.com/eag/esn/blob/main/demo/predict_k20.mp4
• 8 step prediction (16 ms) of broadband eeg activity (0.1 to 40 Hz): https://github.com/eag/esn/blob/main/demo/predict_k8.mp4

Dependencies

• cxxopts (https://github.com/jarro2783/cxxopts)
• armadillo (https://github.com/EmanueleCannizzaro/armadillo)