### Get Expert Help From The Gensim Authors

Consulting in Machine Learning & NLP

Corporate trainings in Data Science, NLP and Deep Learning

_matutils – Cython matutils

# `_matutils` – Cython matutils¶

`gensim._matutils.``digamma`

Digamma function for positive floats, using `_digamma()`.

Parameters

x (float) – Positive value.

Returns

Digamma(x).

Return type

float

`gensim._matutils.``dirichlet_expectation`(alpha)

Expected value of log(theta) where theta is drawn from a Dirichlet distribution. Using `dirichlet_expectation_1d()` or `dirichlet_expectation_2d()`.

Parameters

alpha (numpy.ndarray) – Dirichlet parameter 2d matrix or 1d vector, if 2d - each row is treated as a separate parameter vector, supports float16, float32 and float64.

Returns

Log of expected values, dimension same as alpha.ndim.

Return type

numpy.ndarray

`gensim._matutils.``dirichlet_expectation_1d`(alpha)

Expected value of log(theta) where theta is drawn from a Dirichlet distribution. Using `_dirichlet_expectation_1d()`.

Parameters

alpha (numpy.ndarray) – Dirichlet parameter 1d vector, supports float16, float32 and float64.

Returns

Log of expected values, 1d vector.

Return type

numpy.ndarray

`gensim._matutils.``dirichlet_expectation_2d`(alpha)

Expected value of log(theta) where theta is drawn from a Dirichlet distribution. Using `_dirichlet_expectation_2d()`.

Parameters

alpha (numpy.ndarray) – Dirichlet parameter 2d matrix, each row is treated as a separate parameter vector, supports float16, float32 and float64.

Returns

Log of expected values, 2d matrix.

Return type

numpy.ndarray

`gensim._matutils.``logsumexp`(x)

Log of sum of exponentials, using `_logsumexp_2d()`.

Parameters

x (numpy.ndarray) – Input 2d matrix, supports float16, float32 and float64.

Returns

log of sum of exponentials of elements in x.

Return type

float

Warning

By performance reasons, doesn’t support NaNs or 1d, 3d, etc arrays like `scipy.special.logsumexp()`.

`gensim._matutils.``mean_absolute_difference`(a, b)

Mean absolute difference between two arrays, using `_mean_absolute_difference()`.

Parameters
• a (numpy.ndarray) – Input 1d array, supports float16, float32 and float64.

• b (numpy.ndarray) – Input 1d array, supports float16, float32 and float64.

Returns

mean(abs(a - b)).

Return type

float