Vector Similarity

Overview

Computes the similarity between two given numeric vectors. Unlike other similarity algorithms that operate on graph structure or node properties, this algorithm works directly on user-provided vectors.

Four metrics are supported:

• Cosine: Cosine of the angle between two vectors. See Cosine Similarity for details.
• Pearson: Linear correlation between two vectors. See Pearson Correlation Coefficient for details.
• Euclidean: Normalized Euclidean distance between two vectors. See Euclidean Distance for details.
• Jaccard: Ratio of positions where both vectors are non-zero to positions where at least one is non-zero. See Jaccard Similarity for details.

Parameters

Run Mode

Returns:

Cosine similarity:

GQL
CALL algo.similarityvec({
  vector1: [1.0, 2.0, 3.0],
  vector2: [4.0, 5.0, 6.0]
}) YIELD similarity

Result:

Pearson correlation:

GQL
CALL algo.similarityvec({
  vector1: [1.0, 2.0, 3.0],
  vector2: [4.0, 5.0, 6.0],
  metric: "pearson"
}) YIELD similarity

Result:

Euclidean distance:

GQL
CALL algo.similarityvec({
  vector1: [1.0, 2.0],
  vector2: [4.0, 6.0],
  metric: "euclidean"
}) YIELD similarity

Result:

Jaccard similarity:

GQL
CALL algo.similarityvec({
  vector1: [1.0, 2.0, 3.0],
  vector2: [2.0, 3.0, 4.0],
  metric: "jaccard"
}) YIELD similarity

Result:

Stream Mode

Returns the same columns as run mode, streamed for memory efficiency.

GQL
CALL algo.similarityvec.stream({
  vector1: [1.0, 2.0, 3.0],
  vector2: [4.0, 5.0, 6.0]
}) YIELD similarity
RETURN similarity

Result:

Stats Mode

Returns:

GQL
CALL algo.similarityvec.stats({
  vector1: [1.0, 2.0, 3.0],
  vector2: [4.0, 5.0, 6.0]
}) YIELD similarity

Result: