Similarity score

Supported in: Batch

Returns the similarity score of two embedding vectors.

Expression categories: Distance measurement, Numeric

Declared arguments

• Left embedded vector: The left embedded vector.
  Expression<T>
• Right embedded vector: The right embedded vector.
  Expression<T>
• Similarity metric: The similarity metric for comparing the left and right embeddings.
  Enum<Cosine Distance, Cosine Similarity, Dot Product, Euclidean Distance>

Type variable bounds: T accepts Array<Float>

Output type: Double

Examples

Example 1: Base case

Description: Cosine similarity of the Ada embeddings for the word 'palantir' and 'foundry'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: COSINE_SIMILARITY

---

Example 2: Base case

Description: Cosine similarity between the Ada embeddings for the word 'palantir'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: COSINE_SIMILARITY

---

Example 3: Base case

Description: Dot product of the Ada embeddings for the word 'palantir' and 'foundry'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: DOT_PRODUCT

---

Example 4: Base case

Description: Dot product of the Ada embeddings for the word 'palantir'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: DOT_PRODUCT

---

Example 5: Base case

Description: Euclidean distance between the Ada embeddings for the word 'palantir' and 'foundry'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: EUCLIDEAN_DISTANCE

---

Example 6: Base case

Description: Euclidean distance between the Ada embeddings for the word 'palantir'. Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: EUCLIDEAN_DISTANCE

---

Example 7: Null case

Description: Null inputs should have a null output Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: COSINE_SIMILARITY

---

Example 8: Null case

Description: Null inputs should have a null output Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: DOT_PRODUCT

---

Example 9: Null case

Description: Null inputs should have a null output Argument values:

• Left embedded vector: leftEmbeddedVector
• Right embedded vector: rightEmbeddedVector
• Similarity metric: EUCLIDEAN_DISTANCE

---

Example 10: Edge case

Description: Regular arrays become null when arrays have different length Argument values:

• Left embedded vector: leftArray
• Right embedded vector: rightArray
• Similarity metric: DOT_PRODUCT

---