dictances 1.5.4

Distances and divergences between distributions implemented in python.

Distances and divergences between discrete distributions described as dictionaries implemented in python.

These are meant as fast solutions to compute distances and divergences between discrete distributions, expecially when the two distributions contains a significant amount of events with nill probability which are not described in the dictionaries.

How do I install this package?

As usual, just download it using pip:

pip install dictances

Available metrics

A number of distances and divergences are available:

Distances	Methods
Bhattacharyya distance	bhattacharyya
Bhattacharyya coefficient	bhattacharyya_coefficient
Canberra distance	canberra
Chebyshev distance	chebyshev
Chi Square distance	chi_square
Cosine Distance	cosine
Euclidean distance	euclidean
Hamming distance	hamming
Jensen-Shannon divergence	jensen_shannon
Kullback-Leibler divergence	kullback_leibler
Mean absolute error	mae
Taxicab geometry	manhattan, cityblock, total_variation
Minkowski distance	minkowsky
Mean squared error	mse
Pearson’s distance	pearson
Squared deviations from the mean	squared_variation

Usage example with points

Suppose you have a point described by my_first_dictionary and another one described by my_second_dictionary:

from dictances import cosine

my_first_dictionary = {
    "a": 56,
    "b": 34,
    "c": 89
}

my_second_dictionary = {
    "a": 21,
    "d": 51,
    "e": 74
}

cosine(my_first_dictionary, my_second_dictionary)
#>>> 0.8847005261889619

Usage example with distributions

Suppose you have a point described by my_first_dictionary and another one described by my_second_dictionary:

from dictances import bhattacharyya, bhattacharyya_coefficient

a = {
    "event_1": 0.4,
    "event_2": 0.1,
    "event_3": 0.2,
    "event_4": 0.3,
}
b = {
    "event_1": 0.1,
    "event_2": 0.2,
    "event_5": 0.2,
    "event_9": 0.5,
}

bhattacharyya_coefficient(a, b)
#>>> 0.3414213562373095
bhattacharyya(a, b)
#>>> 1.07463791569453

Handling nested dictionaries

If you need to compute the distance between two nested dictionaries you can use deflate_dict as follows:

from dictances import cosine
from deflate_dict import deflate

my_first_dictionary = {
    "a": 8,
    "b": {
        "c": 3,
        "d": 6
    }
}

my_second_dictionary = {
    "b": {
        "c": 8,
        "d": 1
    },
    "y": 3,

}

cosine(deflate(my_first_dictionary), deflate(my_second_dictionary))