Graph Reasoning with Reinforcement Learning

This repository provides PyTorch implementations of reinforcement learning algorithms for reasoning over knowledge graphs.

Algorithms

• DeepPath (EMNLP 2017): A reinforcement learning method for knowledge graph reasoning that learns to find paths between entities.

• MINERVA (ICLR 2018): Multi-hop reasoning that learns to navigate knowledge graphs by taking a sequence of relation-following steps.

Each algorithm directory contains its own detailed documentation and usage instructions.