Knowledge Graph Embeddings using Neural Ito Process: From Multiple Walks to Stochastic Trajectories
Mojtaba Nayyeri, Bo Xiong, Majid Mohammadi, Mst. Mahfuja Akter, Mirza Mohtashim Alam, Jens Lehmann, Steffen Staab · 2023
Knowledge graphs mostly exhibit a mixture of branching relations, e.g., hasFriend, and complex structures, e.g., hierarchy and loop.Most knowledge graph embeddings have problems expressing them, because they model a specific relation r from a head h to tails by starting at the node embedding of h and transitioning deterministically to exactly one other point in the embedding space.We overcome this issue in our novel framework ItôE by modeling relations between nodes by relation-specific, stochastic transitions.Our framework is based on stochastic Itô processes, which operate on low-dimensional manifolds.ItôE is highly expressive and generic subsuming various stateof-the-art models operating on different, also non-Euclidean, manifolds.Experimental results show the superiority of ItôE over other deterministic embedding models with regard to the KG completion task.