Feature Ordering by Conditional Dependence (FOCI)

FOCI¹ is a forward stepwise feature selection algorithm for multivariate regression based on Conditional Dependence measure.

Usage

Make a random 2000x100 array of independent variables, and simulate a complex Y

import numpy as np


n = 2000
p = 100
x = np.random.normal(0, 1, (n, p))
y = x[:, 0] * x[:, 1] + np.sin(x[:, 2] * x[:, 0])

Select Features

import xicorpy

selected = xicorpy.select_features_using_foci(y, x, num_features=10)

Select Features with Initial Feature Set

import xicorpy

selected = xicorpy.select_features_using_foci(y, x, num_features=10, init_selection=[0])