Relational recurrent neural networks

Abstract

Memory-based neural networks model temporal data by leveraging an ability to remember information for long periods. It is unclear, however, whether they also have an ability to perform complex relational reasoning with the information they remember. Here, we first confirm our intuitions that standard memory architectures may struggle at tasks that heavily involve an understanding of the ways in which entities are connected -- i.e., tasks involving relational reasoning. We then improve upon these deficits by using a new memory module -- a \textit{Relational Memory Core} (RMC) -- which employs multi-head dot product attention to allow memories to interact. Finally, we test the RMC on a suite of tasks that may profit from more capable relational reasoning across sequential information, and show large gains in RL domains (e.g. Mini PacMan), program evaluation, and language modeling, achieving state-of-the-art results on the WikiText-103, Project Gutenberg, and GigaWord datasets.

References in corpus (20)

• Adam: A Method for Stochastic Optimization
• Sequence to Sequence Learning with Neural Networks
• Semi-Supervised Classification with Graph Convolutional Networks
• Neural Message Passing for Quantum Chemistry
• Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation
• Language Modeling with Gated Convolutional Networks
• Exploring the Limits of Language Modeling
• IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
• A simple neural network module for relational reasoning
• Pointer Sentinel Mixture Models
• Learning to Execute
• Non-Local Recurrent Network for Image Restoration
• How to Construct Deep Recurrent Neural Networks
• Tracking the World State with Recurrent Entity Networks
• Non-local Neural Networks
• Discovering objects and their relations from entangled scene representations
• Relation Networks for Object Detection
• Breaking the Softmax Bottleneck: A High-Rank RNN Language Model
• Fast Parametric Learning with Activation Memorization
• Working Memory Networks: Augmenting Memory Networks with a Relational Reasoning Module

Cited by in corpus (11)

• Analysing Mathematical Reasoning Abilities of Neural Models
• A neural network walks into a lab: towards using deep nets as models for human behavior
• Investigating Object Compositionality in Generative Adversarial Networks
• Object Files and Schemata: Factorizing Declarative and Procedural Knowledge in Dynamical Systems
• Adaptive Posterior Learning: few-shot learning with a surprise-based memory module
• Relational Action Forecasting
• Fitted Q-Learning for Relational Domains
• CatGAN: Category-aware Generative Adversarial Networks with Hierarchical Evolutionary Learning for Category Text Generation
• Particle Filter Recurrent Neural Networks
• Perception-Prediction-Reaction Agents for Deep Reinforcement Learning
• Relational dynamic memory networks