metrics

>>> import perceval as pcvl
>>> dist_a = pcvl.BSDistribution({pcvl.BasicState([1, 0]): 0.4, pcvl.BasicState([0, 1]): 0.6})
>>> dist_b = pcvl.BSDistribution({pcvl.BasicState([1, 0]): 0.3, pcvl.BasicState([0, 1]): 0.7})
>>> print(pcvl.tvd_dist(dist_a, dist_b))
0.1
>>> print(pcvl.kl_divergence(dist_a, dist_b))
0.022582421084357485

Perceval provides ways to compare BSDistribution with mathematical metrics.

perceval.utils.dist_metrics.kl_divergence(ideal_dist, est_dist)

Computes the Kullback-Leibler (KL) divergence of a model (simulated/observed) BSDistribution with respect to an ideal BSDistribution. Our computation ignores states absent from the estimated distribution or have null probabilities.

Parameters:

• ideal_dist (BSDistribution) – Ideal BSDistribution (known from theory or an ideal computation)

• est_dist (BSDistribution) – Estimated BSDistribution (simulated or observed from experiment)

Return type:

float

Returns:

KL divergence of the estimated distribution relative to the ideal.

perceval.utils.dist_metrics.tvd_dist(dist_lh, dist_rh)

Computes the Total Variation Distance (TVD) between two input BSDistributions.

Parameters:

• dist_lh (BSDistribution) – First BSDistribution

• dist_rh (BSDistribution) – Second BSDistribution

Return type:

float

Returns:

total variation distance between the two BSDistributions (value between 0 and 1)