Enum Class QdrantDistance
- All Implemented Interfaces:
Serializable,Comparable<QdrantDistance>,Constable
The distance metric determines how Qdrant compares stored vectors with a query vector. The metric is part of the vector collection/index configuration and is not selected per search operation.
A typical configuration may use:
QdrantDistance distance = QdrantDistance.COSINE;
Conceptually:
query vector
|
v
configured distance
|
v
stored vectors
|
v
nearest matches
Each value maps to the equivalent Qdrant client Collections.Distance.
-
Nested Class Summary
Nested classes/interfaces inherited from class java.lang.Enum
Enum.EnumDesc<E extends Enum<E>> -
Enum Constant Summary
Enum Constants -
Method Summary
Modifier and TypeMethodDescriptionstatic QdrantDistanceParses a configuration value, ignoring case and surrounding whitespace.static QdrantDistanceReturns the enum constant of this class with the specified name.static QdrantDistance[]values()Returns an array containing the constants of this enum class, in the order they are declared.
-
Enum Constant Details
-
COSINE
Cosine distance.Cosine focuses mainly on the direction of the vectors rather than their magnitude. This is useful when the relative pattern matters more than the absolute amount.
Consider a recommendation system where vector dimensions represent movie genres:
Action Comedy Drama Alice 10 2 1 Bob 100 20 10 Carol 2 10 8Alice and Bob have very different numbers of watched movies, but their preference proportions are essentially the same. Both strongly prefer action movies over comedy and drama.
Cosine therefore considers Alice and Bob highly similar because their vectors point in the same direction. Carol has a different preference pattern and would be considered less similar.
This behavior is often useful for embeddings and recommendation systems, where the pattern or semantic direction matters more than the absolute magnitude.
A · B cosine(A,B) = ---------- |A| |B| -
EUCLID
Euclidean distance, also known as L2 distance.Euclidean distance measures the straight-line distance between two vectors. Unlike cosine, the magnitude of the vectors matters.
Consider the same movie-preference example:
Action Comedy Drama Alice 10 2 1 Bob 100 20 10 Carol 11 3 1Alice and Bob have essentially the same preference proportions, so cosine would consider them very similar.
Euclidean distance behaves differently because it considers the absolute values. Bob is much farther from Alice because he watched many more movies in every category.
Carol, on the other hand, is very close to Alice because their absolute values are also similar:
Alice [10, 2, 1] Carol [11, 3, 1]
Therefore, Euclidean distance is useful when the magnitude or intensity represented by the vector is meaningful, not only its direction.
euclid(A,B) = sqrt( (A1 - B1)^2 + (A2 - B2)^2 + ... + (An - Bn)^2 ) -
DOT
Dot-product similarity.Dot product considers both direction and magnitude. Two users with the same preference pattern can receive different scores if one has much stronger activity than the other.
Consider:
Action Comedy Drama Query 1 1 1 Alice 10 2 1 Bob 100 20 10 Carol 2 10 8Alice and Bob have the same preference proportions, but Bob has a much larger magnitude because he watched many more movies.
With dot product, Bob produces a much larger score than Alice:
Query · Alice = 1*10 + 1*2 + 1*1 = 13 Query · Bob = 1*100 + 1*20 + 1*10 = 130
This means dot product can be useful when both the preference direction and the strength or intensity represented by the vector matter.
dot(A,B) = A1 * B1 + A2 * B2 + ... + An * Bn -
MANHATTAN
Manhattan distance, also known as L1 distance.Manhattan distance adds the absolute difference between each dimension. Each movie category contributes independently to the final distance.
Consider:
Action Comedy Drama Alice 10 2 1 Bob 100 20 10 Carol 11 3 1Carol is close to Alice because the difference in each category is small:
|10 - 11| + |2 - 3| + |1 - 1| = 1 + 1 + 0 = 2
Bob is much farther away:
|10 - 100| + |2 - 20| + |1 - 10| = 90 + 18 + 9 = 117
Unlike cosine, Manhattan distance cares about the absolute difference in activity. Unlike Euclidean distance, it simply adds the differences instead of combining them using squared geometric distance.
It is useful when each dimension represents an independent difference that should contribute directly to the total distance.
manhattan(A,B) = |A1 - B1| + |A2 - B2| + ... + |An - Bn|
-
-
Method Details
-
values
Returns an array containing the constants of this enum class, in the order they are declared.- Returns:
- an array containing the constants of this enum class, in the order they are declared
-
valueOf
Returns the enum constant of this class with the specified name. The string must match exactly an identifier used to declare an enum constant in this class. (Extraneous whitespace characters are not permitted.)- Parameters:
name- the name of the enum constant to be returned.- Returns:
- the enum constant with the specified name
- Throws:
IllegalArgumentException- if this enum class has no constant with the specified nameNullPointerException- if the argument is null
-
parse
Parses a configuration value, ignoring case and surrounding whitespace. The Qdrant name for the metric (Cosine,Euclid,Dot,Manhattan) is accepted.- Parameters:
value- the configuration value- Returns:
- the matching metric
- Throws:
NullPointerException- when the value is nullIllegalArgumentException- when the value is not a supported metric
-