Learning Neural-Symbolic Descriptive Planning Models via Cube-Space Priors: The Voyage Home (to STRIPS)

Abstract

We achieved a new milestone in the difficult task of enabling agents to learn about their environment autonomously. Our neuro-symbolic architecture is trained end-to-end to produce a succinct and effective discrete state transition model from images alone. Our target representation (the Planning Domain Definition Language) is already in a form that off-the-shelf solvers can consume, and opens the door to the rich array of modern heuristic search capabilities. We demonstrate how the sophisticated innate prior we place on the learning process significantly reduces the complexity of the learned representation, and reveals a connection to the graph-theoretic notion of "cube-like graphs", thus opening the door to a deeper understanding of the ideal properties for learned symbolic representations. We show that the powerful domain-independent heuristics allow our system to solve visual 15-Puzzle instances which are beyond the reach of blind search, without resorting to the Reinforcement Learning approach that requires a huge amount of training on the domain-dependent reward information.

arXiv commentAccepted in IJCAI 2020 main track (accept ratio 12.6%). The prequel of this paper, "The Search for STRIPS", can be found here: arXiv:1912.05492 . (update, 2020/08/11) We expanded the related work section

References in corpus (12)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Categorical Reparameterization with Gumbel-Softmax
• Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computation
• The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables
• Learning Plannable Representations with Causal InfoGAN
• Neural Logic Machines
• DVAE++: Discrete Variational Autoencoders with Overlapping Transformations
• Learning Finite State Representations of Recurrent Policy Networks
• DVAE#: Discrete Variational Autoencoders with Relaxed Boltzmann Priors
• Learning First-Order Symbolic Representations for Planning from the Structure of the State Space
• Learning Action Models from Disordered and Noisy Plan Traces
• Neural-Symbolic Descriptive Action Model from Images: The Search for STRIPS

Cited by in corpus (3)

• DeepSym: Deep Symbol Generation and Rule Learning from Unsupervised Continuous Robot Interaction for Planning
• Goal Recognition as Reinforcement Learning
• Discrete Word Embedding for Logical Natural Language Understanding