Tutorial 4: Standardizing Within Runs Using standardize

While standardizing the features/columns within runs can be done using the standardize parameter within the TimeseriesExtractor class, the standardize() function can be used to perform within-run standardization post-extraction.

import numpy as np
from neurocaps.analysis import standardize

subject_timeseries = {
    str(x): {f"run-{y}": np.random.rand(10, 100) for y in range(1, 4)} for x in range(1, 6)
}

# Getting mean and standard deviation for run 1 and 2 of subject 1
mean_vec_1 = subject_timeseries["1"]["run-1"].mean(axis=0)
std_vec_1 = subject_timeseries["1"]["run-1"].std(ddof=1, axis=0)
mean_vec_2 = subject_timeseries["1"]["run-2"].mean(axis=0)
std_vec_2 = subject_timeseries["1"]["run-2"].std(ddof=1, axis=0)

# Avoid numerical stability issues
std_vec_1[std_vec_1 < np.finfo(std_vec_1.dtype).eps] = 1.0
std_vec_2[std_vec_2 < np.finfo(std_vec_2.dtype).eps] = 1.0

standardized_subject_timeseries = standardize(subject_timeseries_list=[subject_timeseries])

standardized_1 = (subject_timeseries["1"]["run-1"] - mean_vec_1) / std_vec_1
standardized_2 = (subject_timeseries["1"]["run-2"] - mean_vec_2) / std_vec_2

print(np.array_equal(standardized_subject_timeseries["dict_0"]["1"]["run-1"], standardized_1))
print(np.array_equal(standardized_subject_timeseries["dict_0"]["1"]["run-2"], standardized_2))

True
True