Differentiable Learning of Logical Rules for Knowledge Base Reasoning
This addresses the challenge of knowledge base reasoning for AI systems, offering a novel method but appears incremental as it builds on existing differentiable logic approaches.
The paper tackles the problem of learning probabilistic first-order logical rules for knowledge base reasoning by proposing Neural Logic Programming, an end-to-end differentiable model that combines parameter and structure learning. Empirically, it outperforms prior work on benchmark datasets like Freebase and WikiMovies, though no specific numbers are provided.
We study the problem of learning probabilistic first-order logical rules for knowledge base reasoning. This learning problem is difficult because it requires learning the parameters in a continuous space as well as the structure in a discrete space. We propose a framework, Neural Logic Programming, that combines the parameter and structure learning of first-order logical rules in an end-to-end differentiable model. This approach is inspired by a recently-developed differentiable logic called TensorLog, where inference tasks can be compiled into sequences of differentiable operations. We design a neural controller system that learns to compose these operations. Empirically, our method outperforms prior work on multiple knowledge base benchmark datasets, including Freebase and WikiMovies.