TransE

The problem, from first principles

TransE is a knowledge graph embedding model. For each triplet (h, r, t), it learns embeddings such that h + r ≈ t. The energy function penalizes violations of this relation.

From theory to code

Implement:

transe_score(h, r, t)

Computes TransE energy for triplet.

Constraints

• h, r, t shape: (embedding_dim,).
• Return: scalar score (lower is better).