diPLSlib.utils subpackage

diPLSlib.utils.misc module

Some helper functions for diPLSlib

diPLSlib.utils.misc.calibrateAnalyticGaussianMechanism(epsilon, delta, GS, tol=1e-12)[source]

Calibrate a Gaussian perturbation for differential privacy using the analytic Gaussian mechanism of [Balle and Wang, ICML’18]

Parameters:

epsilonfloat: Privacy parameter epsilon
deltafloat: Desired privacy failure probability
GSfloat: Upper bound on the L2-sensitivity of the function to which the mechanism is applied
tolfloat: Error tolerance for binary search

Returns:

sigmafloat: Standard deviation of Gaussian noise needed to achieve (epsilon,delta)-DP under global sensitivity GS

References

Balle, B., & Wang, Y. X. (2018, July). Improving the gaussian mechanism for differential privacy: Analytical calibration and optimal denoising. In International Conference on Machine Learning (pp. 394-403). PMLR.

Examples

>>> from diPLSlib.utils.misc import calibrateAnalyticGaussianMechanism
>>> calibrateAnalyticGaussianMechanism(1.0, 1e-5, 1.0)
3.730631634944469

diPLSlib.utils.misc.gengaus(length, mu, sigma, mag, noise=0)[source]

Generate a Gaussian spectrum-like signal with optional random noise.

Parameters:

lengthint: Length of the generated signal.
mufloat: Mean of the Gaussian function.
sigmafloat: Standard deviation of the Gaussian function.
magfloat: Magnitude of the Gaussian signal.
noisefloat, optional (default=0): Standard deviation of the Gaussian noise to be added to the signal.

Returns:

signalndarray of shape (length,): The generated Gaussian signal with noise.

Examples

>>> from diPLSlib.utils.misc import gengaus
>>> import numpy as np
>>> import scipy.stats
>>> signal = gengaus(100, 50, 10, 5, noise=0.1)

diPLSlib.utils.misc.hellipse(X, alpha=0.05)[source]

Compute the 95% confidence interval ellipse for a 2D scatter plot.

Parameters:

Xndarray of shape (n_samples, 2): Matrix of data points.
alphafloat, optional (default=0.05): Significance level for the confidence interval.

Returns:

elndarray of shape (2, 100): Coordinates of the ellipse’s points. To plot, use plt.plot(el[0, :], el[1, :]).

Examples

>>> import matplotlib.pyplot as plt
>>> import numpy as np
>>> from diPLSlib.utils.misc import hellipse
>>> X = np.random.random((100, 2))
>>> el = hellipse(X)
>>> plt.scatter(X[:,0], X[:,1], label='Data points')            
<matplotlib.collections.PathCollection object at ...>
>>> plt.plot(el[0,:], el[1,:], label='95% Confidence Ellipse')  
[<matplotlib.lines.Line2D object at ...>]
>>> plt.legend()                                                
<matplotlib.legend.Legend object at ...>

diPLSlib.utils.misc.rmse(y, yhat)[source]

Compute the Root Mean Squared Error (RMSE) between two arrays.

Parameters:

yndarray of shape (n_samples,): True values.
yhatndarray of shape (n_samples,): Predicted values.

Returns:

errorndarray of shape (n_samples,): The RMSE between y and yhat.

Examples

>>> import numpy as np
>>> from diPLSlib.utils.misc import rmse
>>> x = np.array([1, 2, 3])
>>> y = np.array([2, 3, 4])
>>> error = rmse(x, y)
>>> print(error)
1.0

diPLSlib.utils subpackage

diPLSlib.utils.misc module

Module contents