def test_full_keys(self): # noqa:D102
# Create an iterator over the 20 subjects.
all_subjects = (self.data[:, :, :, i] for i in range(20))
option = 'full'
# Check the keys are identical.
self.assertEqual(sorted(functional_dimensionality(
all_subjects, 20, self.mask, option=option).keys()),
sorted(output.dictionary_full.keys()))
def test_functional_dimensionality_res(self): # noqa:D102
# Create an iterator over the 20 subjects.
all_subjects = np.asarray([self.data[:, :, :, i] for i in range(20)])
option = 'full'
res = np.random.random(all_subjects.shape)
# Check the keys are identical.
self.assertEqual(sorted(functional_dimensionality(
all_subjects, 20, self.mask, res=res, option=option).keys()),
sorted(output.dictionary_full.keys()))
def test_mean(self): # noqa:D102
for d in range(1, 10):
# Get some randomly generated data with d dimensions as ground
# truth:
data = demo_data(nvoxels=self.nvoxels, nsubs=self.nsubs,
functional_dims=d, nconditions=16)
# Create an iterator over the 20 subjects.
all_subjects = (data[:, :, :, i] for i in range(20))
option = 'mean'
self.assertEqual(d, functional_dimensionality(
all_subjects, self.nsubs, self.mask, option=option)
['winning_model'].mean())
def test_mean_values(self): # noqa:D102
# Create an iterator over the 20 subjects.
all_subjects = (self.data[:, :, :, i] for i in range(20))
option = 'mean'
# Loop through to compare the contents of both dictionaries:
for (key, value) in functional_dimensionality(
all_subjects, self.nsubs, self.mask, option=option).items():
if key == 'subject_ID':
self.assertTrue((value == output.dictionary_mean[key]).all())
else:
self.assertTrue(np.allclose(value,
output.dictionary_mean[key]))
import pandas as pd
# load the sample data.
data = np.load('./demos/demo_data/sample_data.npy')
# "data" has the shape (64, 16, 6, 20), containing beta values for 64 voxels,
# 16 conditions, 6 sessions, 20 subjects.
nsubs = 20
# Create the subject IDs.
subject_IDs = [str(i) for i in range(1, nsubs + 1)]
# Create a 4*4*4 mask (all True) for the 64 voxels.
mask = np.ones((4, 4, 4), dtype='bool')
# Create an iterator over the 20 subjects.
all_subjects = (data[:, :, :, i] for i in range(nsubs))
# Find the dimensionality.
results = functional_dimensionality(all_subjects,
nsubs,
mask,
subject_IDs=subject_IDs,
option='mean')
# Put the output results into a dataframe and save as a CSV file.
df = pd.DataFrame.from_dict(results)
df.to_csv('./demos/demo_data/demo_real_data_output.csv')
# Show the median dimensionality:
print(df['winning_model'].median())
#!/usr/bin/env python3
from funcdim.funcdim import functional_dimensionality
import numpy as np
# load the sample data.
data = np.load('demo_data/sample_data.npy')
# "data" has the shape (64, 16, 6, 20), containing beta values for 64 voxels,
# 16 conditions, 6 sessions, 20 subjects.
# Create a 4*4*4 mask (all True) for the 64 voxels.
mask = np.ones((4, 4, 4), dtype='bool')
# Create an iterator over the 20 subjects.
all_subjects = (data[:, :, :, i] for i in range(20))
# Find the dimensionality.
results = functional_dimensionality(all_subjects, 20, mask)
print(results)