Timezone: »
We study the problem of inferring communication structures that can solve cooperative multi-agent planning problems while minimizing the amount of communication. We quantify the amount of communication as the maximum degree of the communication graph; this metric captures settings where agents have limited bandwidth. Minimizing communication is challenging due to the combinatorial nature of both the decision space and the objective; for instance, we cannot solve this problem by training neural networks using gradient descent. We propose a novel algorithm that synthesizes a control policy that combines a programmatic communication policy used to generate the communication graph with a transformer policy network used to choose actions. Our algorithm first trains the transformer policy, which implicitly generates a "soft" communication graph; then, it synthesizes a programmatic communication policy that "hardens" this graph, forming a neurosymbolic transformer. Our experiments demonstrate how our approach can synthesize policies that generate low-degree communication graphs while maintaining near-optimal performance.
Author Information
Jeevana Priya Inala (MIT)
Yichen Yang (MIT)
James Paulos (University of Pennsylvania)
Yewen Pu (autodesk)
Osbert Bastani (University of Pennysylvania)
Vijay Kumar (University of Pennsylvania)
Martin Rinard (MIT)
Armando Solar-Lezama (MIT)
More from the Same Authors
-
2020 Workshop: Workshop on Computer Assisted Programming (CAP) »
Augustus Odena · Charles Sutton · Nadia Polikarpova · Josh Tenenbaum · Armando Solar-Lezama · Isil Dillig -
2020 Poster: Program Synthesis with Pragmatic Communication »
Yewen Pu · Kevin Ellis · Marta Kryven · Josh Tenenbaum · Armando Solar-Lezama -
2020 Poster: Learning Compositional Rules via Neural Program Synthesis »
Maxwell Nye · Armando Solar-Lezama · Josh Tenenbaum · Brenden Lake -
2020 Poster: Efficient Exact Verification of Binarized Neural Networks »
Kai Jia · Martin Rinard -
2019 Poster: Write, Execute, Assess: Program Synthesis with a REPL »
Kevin Ellis · Maxwell Nye · Yewen Pu · Felix Sosa · Josh Tenenbaum · Armando Solar-Lezama -
2019 Poster: Compiler Auto-Vectorization with Imitation Learning »
Charith Mendis · Cambridge Yang · Yewen Pu · Saman Amarasinghe · Michael Carbin -
2019 Poster: A Composable Specification Language for Reinforcement Learning Tasks »
Kishor Jothimurugan · Rajeev Alur · Osbert Bastani -
2018 Poster: Learning to Infer Graphics Programs from Hand-Drawn Images »
Kevin Ellis · Daniel Ritchie · Armando Solar-Lezama · Josh Tenenbaum -
2018 Poster: Learning Libraries of Subroutines for Neurally–Guided Bayesian Program Induction »
Kevin Ellis · Lucas Morales · Mathias Sablé-Meyer · Armando Solar-Lezama · Josh Tenenbaum -
2018 Spotlight: Learning to Infer Graphics Programs from Hand-Drawn Images »
Kevin Ellis · Daniel Ritchie · Armando Solar-Lezama · Josh Tenenbaum -
2018 Spotlight: Learning Libraries of Subroutines for Neurally–Guided Bayesian Program Induction »
Kevin Ellis · Lucas Morales · Mathias Sablé-Meyer · Armando Solar-Lezama · Josh Tenenbaum -
2018 Poster: Verifiable Reinforcement Learning via Policy Extraction »
Osbert Bastani · Yewen Pu · Armando Solar-Lezama -
2018 Poster: Interpreting Neural Network Judgments via Minimal, Stable, and Symbolic Corrections »
Xin Zhang · Armando Solar-Lezama · Rishabh Singh -
2016 Poster: Sampling for Bayesian Program Learning »
Kevin Ellis · Armando Solar-Lezama · Josh Tenenbaum -
2015 Poster: Unsupervised Learning by Program Synthesis »
Kevin Ellis · Armando Solar-Lezama · Josh Tenenbaum