striatum
Reputation: 1598
Python `sklearn.metrics.pairwise` `cosine_similarity` or `cosine_distance` theoretical range
Cosine similarity/distance should be bounded between -1 and +1. Yet, sklearn.metrics.pairwise gives values > +1. Can anyone explain this behaviour please?
Upvotes: 0
Views: 506
Answers (1)
Learning is a mess
Reputation: 8277
pairwise default distance is Euclidean which is not bounded from above, hence you should not be surpised to see some values greater than 1.
Upvotes: 1
Related Questions
- Difference between cosine similarity and cosine distance
- Efficient (not DataFrame.apply) way of getting cosine distance for mapped values
- Cosine distance more than 1
- cosine_similarity between 2 pandas df column to get cosine distance
- What is the best way to implement an element-wise cosine similarity in Python?
- scikit cosine_similarity vs pairwise_distances
- calculate cosine similarity matrix without scipy, sklearn.metrics.pairwise
- What is the cosine distance of something and itself?
- Cosine Similarity
- Distance metrics in Scikit Learn