Timezone: »
Learning in general-sum games is unstable and frequently leads to socially undesirable (Pareto-dominated) outcomes. To mitigate this, Learning with Opponent-Learning Awareness (LOLA) introduced opponent shaping to this setting, by accounting for each agent's influence on their opponents' anticipated learning steps. However, the original LOLA formulation (and follow-up work) is inconsistent because LOLA models other agents as naive learners rather than LOLA agents.In previous work, this inconsistency was suggested as a cause of LOLA's failure to preserve stable fixed points (SFPs). First, we formalize consistency and show that higher-order LOLA (HOLA) solves LOLA's inconsistency problem if it converges. Second, we correct a claim made in the literature by Schäfer and Anandkumar (2019), proving that Competitive Gradient Descent (CGD) does not recover HOLA as a series expansion (and fails to solve the consistency problem).Third, we propose a new method called Consistent LOLA (COLA), which learns update functions that are consistent under mutual opponent shaping. It requires no more than second-order derivatives and learns consistent update functions even when HOLA fails to converge. However, we also prove that even consistent update functions do not preserve SFPs, contradicting the hypothesis that this shortcoming is caused by LOLA's inconsistency.Finally, in an empirical evaluation on a set of general-sum games, we find that COLA finds prosocial solutions and that it converges under a wider range of learning rates than HOLA and LOLA. We support the latter finding with a theoretical result for a simple game.
Author Information
Timon Willi (University of Oxford)
Alistair Letcher (None)
Johannes Treutlein (University of Toronto, Vector Institute)
Jakob Foerster (Oxford university)
Jakob Foerster started as an Associate Professor at the department of engineering science at the University of Oxford in the fall of 2021. During his PhD at Oxford he helped bring deep multi-agent reinforcement learning to the forefront of AI research and interned at Google Brain, OpenAI, and DeepMind. After his PhD he worked as a research scientist at Facebook AI Research in California, where he continued doing foundational work. He was the lead organizer of the first Emergent Communication workshop at NeurIPS in 2017, which he has helped organize ever since and was awarded a prestigious CIFAR AI chair in 2019. His past work addresses how AI agents can learn to cooperate and communicate with other agents, most recently he has been developing and addressing the zero-shot coordination problem setting, a crucial step towards human-AI coordination.
Related Events (a corresponding poster, oral, or spotlight)
-
2022 Poster: COLA: Consistent Learning with Opponent-Learning Awareness »
Thu. Jul 21st through Fri the 22nd Room Hall E #811
More from the Same Authors
-
2022 : Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2022 : Illusionary Attacks on Sequential Decision Makers and Countermeasures »
Tim Franzmeyer · Joao Henriques · Jakob Foerster · Phil Torr · Adel Bibi · Christian Schroeder -
2022 : Discovered Policy Optimisation »
Christopher Lu · Jakub Grudzien Kuba · Alistair Letcher · Luke Metz · Christian Schroeder · Jakob Foerster -
2022 : Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2022 : Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2022 : Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2023 : Illusory Attacks: Detectability Matters in Adversarial Attacks on Sequential Decision-Makers »
Tim Franzmeyer · Stephen Mcaleer · Joao Henriques · Jakob Foerster · Phil Torr · Adel Bibi · Christian Schroeder -
2023 : Analyzing the Sample Complexity of Model-Free Opponent Shaping »
Kitty Fung · Qizhen Zhang · Christopher Lu · Timon Willi · Jakob Foerster -
2023 : Structured State Space Models for In-Context Reinforcement Learning »
Christopher Lu · Yannick Schroecker · Albert Gu · Emilio Parisotto · Jakob Foerster · Satinder Singh · Feryal Behbahani -
2023 : Who to imitate: Imitating desired behavior from diverse multi-agent datasets »
Tim Franzmeyer · Jakob Foerster · Edith Elkind · Phil Torr · Joao Henriques -
2023 Poster: Learning Intuitive Policies Using Action Features »
Mingwei Ma · Jizhou Liu · Samuel Sokota · Max Kleiman-Weiner · Jakob Foerster -
2023 Poster: Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2022 : Adversarial Cheap Talk »
Christopher Lu · Timon Willi · Alistair Letcher · Jakob Foerster -
2022 Poster: Evolving Curricula with Regret-Based Environment Design »
Jack Parker-Holder · Minqi Jiang · Michael Dennis · Mikayel Samvelyan · Jakob Foerster · Edward Grefenstette · Tim Rocktäschel -
2022 Spotlight: Evolving Curricula with Regret-Based Environment Design »
Jack Parker-Holder · Minqi Jiang · Michael Dennis · Mikayel Samvelyan · Jakob Foerster · Edward Grefenstette · Tim Rocktäschel -
2022 Poster: Communicating via Markov Decision Processes »
Samuel Sokota · Christian Schroeder · Maximilian Igl · Luisa Zintgraf · Phil Torr · Martin Strohmeier · Zico Kolter · Shimon Whiteson · Jakob Foerster -
2022 Spotlight: Communicating via Markov Decision Processes »
Samuel Sokota · Christian Schroeder · Maximilian Igl · Luisa Zintgraf · Phil Torr · Martin Strohmeier · Zico Kolter · Shimon Whiteson · Jakob Foerster -
2022 Poster: Model-Free Opponent Shaping »
Christopher Lu · Timon Willi · Christian Schroeder de Witt · Jakob Foerster -
2022 Poster: Mirror Learning: A Unifying Framework of Policy Optimisation »
Jakub Grudzien Kuba · Christian Schroeder de Witt · Jakob Foerster -
2022 Poster: Generalized Beliefs for Cooperative AI »
Darius Muglich · Luisa Zintgraf · Christian Schroeder de Witt · Shimon Whiteson · Jakob Foerster -
2022 Spotlight: Generalized Beliefs for Cooperative AI »
Darius Muglich · Luisa Zintgraf · Christian Schroeder de Witt · Shimon Whiteson · Jakob Foerster -
2022 Spotlight: Model-Free Opponent Shaping »
Christopher Lu · Timon Willi · Christian Schroeder de Witt · Jakob Foerster -
2022 Spotlight: Mirror Learning: A Unifying Framework of Policy Optimisation »
Jakub Grudzien Kuba · Christian Schroeder de Witt · Jakob Foerster -
2021 Poster: Off-Belief Learning »
Hengyuan Hu · Adam Lerer · Brandon Cui · Luis Pineda · Noam Brown · Jakob Foerster -
2021 Spotlight: Off-Belief Learning »
Hengyuan Hu · Adam Lerer · Brandon Cui · Luis Pineda · Noam Brown · Jakob Foerster -
2021 Poster: Trajectory Diversity for Zero-Shot Coordination »
Andrei Lupu · Brandon Cui · Hengyuan Hu · Jakob Foerster -
2021 Spotlight: Trajectory Diversity for Zero-Shot Coordination »
Andrei Lupu · Brandon Cui · Hengyuan Hu · Jakob Foerster -
2021 Poster: A New Formalism, Method and Open Issues for Zero-Shot Coordination »
Johannes Treutlein · Michael Dennis · Caspar Oesterheld · Jakob Foerster -
2021 Spotlight: A New Formalism, Method and Open Issues for Zero-Shot Coordination »
Johannes Treutlein · Michael Dennis · Caspar Oesterheld · Jakob Foerster -
2020 Poster: “Other-Play” for Zero-Shot Coordination »
Hengyuan Hu · Alexander Peysakhovich · Adam Lerer · Jakob Foerster -
2019 Poster: Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning »
Jakob Foerster · Francis Song · Edward Hughes · Neil Burch · Iain Dunning · Shimon Whiteson · Matthew Botvinick · Michael Bowling -
2019 Oral: Bayesian Action Decoder for Deep Multi-Agent Reinforcement Learning »
Jakob Foerster · Francis Song · Edward Hughes · Neil Burch · Iain Dunning · Shimon Whiteson · Matthew Botvinick · Michael Bowling -
2019 Poster: A Baseline for Any Order Gradient Estimation in Stochastic Computation Graphs »
Jingkai Mao · Jakob Foerster · Tim Rocktäschel · Maruan Al-Shedivat · Gregory Farquhar · Shimon Whiteson -
2019 Oral: A Baseline for Any Order Gradient Estimation in Stochastic Computation Graphs »
Jingkai Mao · Jakob Foerster · Tim Rocktäschel · Maruan Al-Shedivat · Gregory Farquhar · Shimon Whiteson -
2018 Poster: The Mechanics of n-Player Differentiable Games »
David Balduzzi · Sebastien Racaniere · James Martens · Jakob Foerster · Karl Tuyls · Thore Graepel -
2018 Poster: QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning »
Tabish Rashid · Mikayel Samvelyan · Christian Schroeder · Gregory Farquhar · Jakob Foerster · Shimon Whiteson -
2018 Oral: The Mechanics of n-Player Differentiable Games »
David Balduzzi · Sebastien Racaniere · James Martens · Jakob Foerster · Karl Tuyls · Thore Graepel -
2018 Oral: QMIX: Monotonic Value Function Factorisation for Deep Multi-Agent Reinforcement Learning »
Tabish Rashid · Mikayel Samvelyan · Christian Schroeder · Gregory Farquhar · Jakob Foerster · Shimon Whiteson -
2018 Poster: DiCE: The Infinitely Differentiable Monte Carlo Estimator »
Jakob Foerster · Gregory Farquhar · Maruan Al-Shedivat · Tim Rocktäschel · Eric Xing · Shimon Whiteson -
2018 Oral: DiCE: The Infinitely Differentiable Monte Carlo Estimator »
Jakob Foerster · Gregory Farquhar · Maruan Al-Shedivat · Tim Rocktäschel · Eric Xing · Shimon Whiteson -
2017 Poster: Stabilising Experience Replay for Deep Multi-Agent Reinforcement Learning »
Jakob Foerster · Nantas Nardelli · Gregory Farquhar · Triantafyllos Afouras · Phil Torr · Pushmeet Kohli · Shimon Whiteson -
2017 Talk: Stabilising Experience Replay for Deep Multi-Agent Reinforcement Learning »
Jakob Foerster · Nantas Nardelli · Gregory Farquhar · Triantafyllos Afouras · Phil Torr · Pushmeet Kohli · Shimon Whiteson -
2017 Poster: Input Switched Affine Networks: An RNN Architecture Designed for Interpretability »
Jakob Foerster · Justin Gilmer · Jan Chorowski · Jascha Sohl-Dickstein · David Sussillo -
2017 Talk: Input Switched Affine Networks: An RNN Architecture Designed for Interpretability »
Jakob Foerster · Justin Gilmer · Jan Chorowski · Jascha Sohl-Dickstein · David Sussillo