Enum RelationalDistanceMetric

Namespace

AiDotNet.KnowledgeDistillation.Strategies

Assembly

AiDotNet.dll

Distance metrics for relational knowledge distillation.

public enum RelationalDistanceMetric

Fields

Cosine = 1

Cosine distance (1 - cosine similarity) - good for high-dimensional embeddings.

Euclidean = 0

Euclidean (L2) distance - most common, geometrically intuitive.

Manhattan = 2

Manhattan (L1) distance - more robust to outliers.