PCA snippet

pca_demo_snip.py

"""
This snippet may be used after installing the following libraries:

# (1) OpenDP library with DP-PCA 
#   - reference: https://test.pypi.org/project/opendp/0.9.0.dev20231221001/
#
pip install -i https://test.pypi.org/simple/ opendp==0.9.0.dev20231221001

# (2) scikit-learn (includes numpy)
#
pip install scikit-learn 

"""

import numpy as np
import opendp.prelude as dp

dp.enable_features("honest-but-curious", "contrib", "floating-point")

# DATASET CREATION
def sample_microdata(*, num_columns=None, num_rows=None, cov=None):
    cov = cov or sample_covariance(num_columns)
    microdata = np.random.multivariate_normal(
        np.zeros(cov.shape[0]), cov, size=num_rows or 100_000
    )
    microdata -= microdata.mean(axis=0)
    return microdata


def sample_covariance(num_features):
    A = np.random.uniform(0, num_features, size=(num_features, num_features))
    return A.T @ A


# USAGE EXAMPLE
num_columns = 4
num_rows = 10_000

model = dp.PCA(
    epsilon=1.,
    row_norm=1.,
    n_samples=num_rows,
    n_features=4,
)

model.fit(sample_microdata(num_columns=num_columns, num_rows=num_rows))

# EXTRACT RELEASES
print("mean", model.mean_)
print("singular values", model.singular_values_)
print("components", model.components_)

loadings = model.singular_values_ * model.components_
print("loadings", loadings)