Metrics

metrics.cosine_similarity(vector1, vector2)

Metric to calcualte the cosine similarity between two vectors.

For example used for the SimilarityTask.

Parameters

• vector1 – input vector 1

• vector2 – input vector 2

Returns

cosine similarity between input vectors between -1 and 1

metrics.mean_cosine_similarity(vectors)

Metric to compute the mean cosine similarity of the given vectors.

For example used for the AnalogyTask.

Parameters

vectors – input vectors

Returns

mean cosine similarity

metrics.mean_pairwise_distance(vectors)

Metric to compute the mean pairwise distance.

For large input sizes use the mean_squared_pairwise_distance.

For example used for the Neighborhood Task.

Parameters

vectors – input vectors

Returns

mean pairwise distance

metrics.mean_squared_pairwise_distance(vectors)

Metric to compute the mean squared pairwise distance. It is equivalent to calculating the variance with n-1.

For example used for the Neighborhood Task.

Parameters

vectors – input vectors

Returns

mean squared pairwise distance

metrics.metric(func)

metrics.metrics = {'cosine_similarity': <function cosine_similarity>, 'mean_cosine_similarity': <function mean_cosine_similarity>, 'mean_pairwise_distance': <function mean_pairwise_distance>, 'mean_squared_pairwise_distance': <function mean_squared_pairwise_distance>}