tick.metrics.support_fdp

tick.metrics.support_fdp(x_truth, x, eps=1e-08)[source]

Computes the False Discovery Proportion for selecting the support of x_truth using x, namely the proportion of false positive among all detected positives, given by FP / (FP + TP). This is useful to assess the features selection or outliers detection abilities of a learner.

Parameters

x_truth : numpy.array

Ground truth weights

x : numpy.array

Learned weights

Returns

output : float

The False Discovery Proportion for detecting the support of x_truth using the support of x

Examples using tick.metrics.support_fdp