TensorLog: Deep Learning Meets Probabilistic DBs

TL;DR

TensorLog integrates probabilistic first-order logic with deep learning frameworks, enabling scalable reasoning and parameter tuning in neural network infrastructures like TensorFlow, demonstrated on large-scale knowledge bases.

Contribution

It introduces a differentiable implementation of probabilistic logic that can be optimized within deep learning frameworks, bridging logical reasoning and neural networks.

Findings

Scales to hundreds of thousands of knowledge-base triples

Handles tens of thousands of examples efficiently

Enables neural network-based parameter tuning for probabilistic logic

Abstract

We present an implementation of a probabilistic first-order logic called TensorLog, in which classes of logical queries are compiled into differentiable functions in a neural-network infrastructure such as Tensorflow or Theano. This leads to a close integration of probabilistic logical reasoning with deep-learning infrastructure: in particular, it enables high-performance deep learning frameworks to be used for tuning the parameters of a probabilistic logic. Experimental results show that TensorLog scales to problems involving hundreds of thousands of knowledge-base triples and tens of thousands of examples.