Recurrent Relational Networks

Abstract

This paper is concerned with learning to solve tasks that require a chain of interdependent steps of relational inference, like answering complex questions about the relationships between objects, or solving puzzles where the smaller elements of a solution mutually constrain each other. We introduce the recurrent relational network, a general purpose module that operates on a graph representation of objects. As a generalization of Santoro et al. [2017]'s relational network, it can augment any neural network model with the capacity to do many-step relational reasoning. We achieve state of the art results on the bAbI textual question-answering dataset with the recurrent relational network, consistently solving 20/20 tasks. As bAbI is not particularly challenging from a relational reasoning point of view, we introduce Pretty-CLEVR, a new diagnostic dataset for relational reasoning. In the Pretty-CLEVR set-up, we can vary the question to control for the number of relational reasoning steps that are required to obtain the answer. Using Pretty-CLEVR, we probe the limitations of multi-layer perceptrons, relational and recurrent relational networks. Finally, we show how recurrent relational networks can learn to solve Sudoku puzzles from supervised training data, a challenging task requiring upwards of 64 steps of relational reasoning. We achieve state-of-the-art results amongst comparable methods by solving 96.6% of the hardest Sudoku puzzles.

Accepted at NIPS 2018

References in corpus (6)

• Neural Message Passing for Quantum Chemistry
• Loopy Belief Propagation for Approximate Inference: An Empirical Study
• A simple neural network module for relational reasoning
• Neural-Symbolic Learning and Reasoning: A Survey and Interpretation
• OptNet: Differentiable Optimization as a Layer in Neural Networks
• Finding ReMO (Related Memory Object): A Simple Neural Architecture for Text based Reasoning

Cited by in corpus (9)

• Learning a SAT Solver from Single-Bit Supervision
• SATNet: Bridging deep learning and logical reasoning using a differentiable satisfiability solver
• Goal-Aware Neural SAT Solver
• Multitask Learning on Graph Neural Networks: Learning Multiple Graph Centrality Measures with a Unified Network
• Deep Reasoning Networks: Thinking Fast and Slow
• Relational Action Forecasting
• From Shallow to Deep Interactions Between Knowledge Representation, Reasoning and Machine Learning (Kay R. Amel group)
• Sentence Encoding with Tree-constrained Relation Networks
• Multi-layer Relation Networks