Knowledge Graph Completion with Mixed Geometry Tensor Factorization
This work addresses knowledge graph completion, a key problem in AI for representing and reasoning with structured data, with incremental improvements in efficiency and performance.
The authors tackled knowledge graph completion by proposing a mixed geometry tensor factorization approach that combines Euclidean and hyperbolic components, achieving new state-of-the-art link prediction accuracy with significantly fewer parameters than previous models.
In this paper, we propose a new geometric approach for knowledge graph completion via low rank tensor approximation. We augment a pretrained and well-established Euclidean model based on a Tucker tensor decomposition with a novel hyperbolic interaction term. This correction enables more nuanced capturing of distributional properties in data better aligned with real-world knowledge graphs. By combining two geometries together, our approach improves expressivity of the resulting model achieving new state-of-the-art link prediction accuracy with a significantly lower number of parameters compared to the previous Euclidean and hyperbolic models.