similarity()

audmath.similarity(u, v)

Cosine similarity between two arrays.

If the incoming arrays are of size $(k,)$, a single similarity value is returned. If one of the incoming arrays is of size $(n, k)$, an array of size $(n,)$ with similarities is returned. If the arrays are of size $(n, k)$ and $(m, k)$ an array of size $(n, m)$ with similarities is returned.

The input arrays can also be provided as pandas.DataFrame or pandas.Series.

The cosine similarity is given by $\frac{u \cdot v}{\lVert u\rVert_2 \lVert v\rVert_2}$.

Parameters

• u (Union[Sequence, ndarray]) – input array

• v (Union[Sequence, ndarray]) – input array

Return type

Union[floating, ndarray]

Returns

similarity between arrays

Example

>>> similarity([1, 0], [1, 0])
np.float64(1.0)
>>> similarity([1, 0], [0, 1])
np.float64(0.0)
>>> similarity([1, 0], [-1, 0])
np.float64(-1.0)
>>> similarity([[1, 0]], [1, 0])
array([1.])
>>> similarity([1, 0], [[1, 0], [0, 1]])
array([1., 0.])
>>> similarity([[1, 0], [0, 1]], [[1, 0]])
array([[1.],
       [0.]])
>>> similarity([[1, 0], [0, 1]], [[1, 0], [0, 1], [-1, 0]])
array([[ 1.,  0., -1.],
       [ 0.,  1.,  0.]])