functions_distance.py

import numpy as np

def norm_data(data):
    m = np.mean(data)
    return np.array([x/m for x in data])


def compute_probs(data, n=10): 
    h, e = np.histogram(data, n)
    p = h/data.shape[0]
    return e, p

def support_intersection(p, q): 
    sup_int = (
        list(
            filter(
                lambda x: (x[0]!=0) & (x[1]!=0), zip(p, q)
            )
        )
    )
    return sup_int

def get_probs(list_of_tuples): 
    p = np.array([p[0] for p in list_of_tuples])
    q = np.array([p[1] for p in list_of_tuples])
    return p, q

def get_probabilities(train_sample, test_sample, n_bins=10):
    
    e, p = compute_probs(train_sample, n=n_bins)
    _, q = compute_probs(test_sample, n=e)

    list_of_tuples = support_intersection(p,q)
    p, q = get_probs(list_of_tuples)

    return p, q
    

def kl_divergence(p, q): 
    return np.sum(p*np.log(p/q))

def js_divergence(p, q):
    m = (1./2.)*(p + q)
    return (1./2.)*kl_divergence(p, m) + (1./2.)*kl_divergence(q, m)