Action Schema Networks: Generalised Policies with Deep Learning

2017/09/13 by Sam Toyer, Felipe Trevizan, Toyer, Sam +5 · 4 citations
Computer Science · #AI-based Problem Solving and Planning #Artificial Intelligence (cs.AI) #Bayesian Modeling and Causal Inference #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.1709.04271

openalex publication_date 2017/09/13 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

In this paper, we introduce the Action Schema Network (ASNet): a neural network architecture for learning generalised policies for probabilistic planning problems. By mimicking the relational structure of planning problems, ASNets are able to adopt a weight-sharing scheme which allows the network to be applied to any problem from a given planning domain. This allows the cost of training the network to be amortised over all problems in that domain. Further, we propose a training method which balances exploration and supervised training on small problems to produce a policy which remains robust when evaluated on larger problems. In experiments, we show that ASNet's learning capability allows it to significantly outperform traditional non-learning planners in several challenging domains.

Cited by

• Learning General Policies From Examples
• Weisfeiler-Leman Features for Planning: A 1,000,000 Sample Size Hyperparameter Study
• Language Models For Generalised PDDL Planning: Synthesising Sound and Programmatic Policies
• LOOP: A Plug-and-Play Neuro-Symbolic Framework for Enhancing Planning in Autonomous Systems

Related

• Learning Generalized Reactive Policies using Deep Neural Networks
• Generalized Planning With Deep Reinforcement Learning
• Graph Neural Network Based Action Ranking for Planning
• What Planning Problems Can A Relational Neural Network Solve?
• Learning General Policies From Examples