Neural NID Rules
This addresses the challenge of generalization in model-based reinforcement learning for AI systems, though it appears incremental as it builds on the classic NID rules framework.
The paper tackled the problem of standard machine learning models in model-based reinforcement learning being inadequate for generalizing to novel situations governed by familiar physics, by introducing Neural NID, a method that learns abstract object properties and relations with a regularized graph neural network, and validated its greater generalization capability on simple benchmarks.
Abstract object properties and their relations are deeply rooted in human common sense, allowing people to predict the dynamics of the world even in situations that are novel but governed by familiar laws of physics. Standard machine learning models in model-based reinforcement learning are inadequate to generalize in this way. Inspired by the classic framework of noisy indeterministic deictic (NID) rules, we introduce here Neural NID, a method that learns abstract object properties and relations between objects with a suitably regularized graph neural network. We validate the greater generalization capability of Neural NID on simple benchmarks specifically designed to assess the transition dynamics learned by the model.