Run ICA On Toy Data

import matplotlib.pyplot as plt
import numpy as np

from sklearn.decomposition import FastICA

import amica
from amica import AMICA

Generate Data and Load AMICA Results for Comparison

%%

data_dir = amica.datasets.data_path() / "toy_2" / "amicaout_toy_2"

Downloading data from 'https://github.com/scott-huberty/amica/releases/download/v0.6.0/test_output.tar.gz' to file '/home/circleci/amica_test_data/74cd649d03dc20d7b938945e84c7c5cd-test_output.tar.gz'.

  0%|                                              | 0.00/4.82M [00:00<?, ?B/s]
  0%|                                              | 0.00/4.82M [00:00<?, ?B/s]
100%|█████████████████████████████████████| 4.82M/4.82M [00:00<00:00, 15.5GB/s]
Untarring contents of '/home/circleci/amica_test_data/74cd649d03dc20d7b938945e84c7c5cd-test_output.tar.gz' to '/home/circleci/amica_test_data/.'

x = amica.utils.generate_toy_data(n_samples=10_000, noise_factor=.05, seed=42)

Run AMICA and FastICA for comparison

fi = FastICA()
z = fi.fit_transform(x)

transformer = AMICA(mean_center=False, whiten="variance", random_state=42,)
transformer.fit(x.copy())

Finished in 8.55 seconds

AMICA(mean_center=False, random_state=42, whiten='variance')

In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook.
On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.

apply the learned unmixing matrix to the data

y = transformer.transform(x)

fortran_results = amica.utils.load_fortran_results(
    data_dir, n_components=2, n_mixtures=3
    )
W_f = fortran_results["W"][:, :, 0]
y2 = x @ (W_f @ fortran_results["S"]).T

Plot Results

fig, ax = plt.subplots(4, 1, sharex=True)
for i, l, v in zip([1, 2, 3, 4], ['Sources', 'AMICA', 'AMICA-Python', 'FastICA'], [x, y2, y, z]):
    ax = plt.subplot(4, 1, i)
    ax.plot(v[:1000])
    ax.set_xticks([])
    ax.set_yticks([])
    ax.set_ylabel(l)
    ax.set_ylim([-1.1 * np.max(v), 1.1 * np.max(v)])