DistanceMetric

DistanceMetric

Bases: StrEnum

Enum for different distance metrics.

Members

- L1_MANHATTAN:            L1 (Manhattan) distance                   = sum_i |x_i - y_i|
- L2_EUCLIDEAN:            L2 (Euclidean) distance                   = sqrt( sum_i (x_i - y_i)^2 )
- L2S_EUCLIDEAN_SQUARED:   L2 squared (Euclidean squared) distance   = sum_i (x_i - y_i)^2
                                  --> avoids computing square root
                                  --> and produces identical solutions for GEOMEAN_SEPARATION diversity metric
- COSINE:                  cosine distance                           = 1 - (x . y) / (|x| |y|)
                                  --> range [0, 2]; angular, for embedding-style vectors
                                  --> undefined for zero vectors (rejected with an error)