Invited Talk
in
Workshop: Machine Learning for Systems

Natasha Jaques: Multi-Agent Reinforcement Learning for Systems

Natasha Jaques

2024 Invited Talk
in
Workshop: Machine Learning for Systems

Abstract

This talk will present new techniques for leveraging deep multi-agent reinforcement learning (MARL) to address NP-hard combinatorial optimization challenges, including the Traveling Salesman Problem (TSP) and Sequential Assignment Problem (SAP). These problems are relevant to a variety of real world systems, ranging from power grids, to satellite assignment, to transportation logistics. Traditional polynomial-time solutions to these problems are computationally expensive and fail to adapt to dynamic real-world conditions, such as the need for rapid re-routing due to unexpected events. Instead, we develop new techniques combining neural network-based reinforcement learning methods which bootstrap from traditional polynomial time solvers. For TSP, we use a genetic curriculum learning method to generate new problems for agents to solve, improving worst-case performance. For SAP, a decentralized MARL approach optimizes task assignments unfolding over time, while initializing estimates using greedy one-step solutions, significantly enhancing efficiency. These techniques have the potential to significantly reduce costs and improve efficiency across a range of systems use cases which require dynamic solutions to routing and assignment problems.

Speaker

Natasha Jaques

Natasha Jaques is an Assistant Professor of Computer Science and Engineering at the University of Washington, and a Senior Research Scientist at Google DeepMind. Her research focuses on Social Reinforcement Learning in multi-agent and human-AI interactions. During her PhD at MIT, she developed techniques for learning from human feedback signals to train language models which were later built on by OpenAI’s series of work on Reinforcement Learning from Human Feedback (RLHF). In the multi-agent space, she has developed techniques for improving coordination through the optimization of social influence, and adversarial environment generation for improving the robustness of RL agents. Natasha’s work has received various awards, including Best Demo at NeurIPS, an honourable mention for Best Paper at ICML, and the Outstanding PhD Dissertation Award from the Association for the Advancement of Affective Computing. Her work has been featured in Science Magazine, MIT Technology Review, Quartz, IEEE Spectrum, Boston Magazine, and on CBC radio, among others. Natasha earned her Masters degree from the University of British Columbia, undergraduate degrees in Computer Science and Psychology from the University of Regina, and completed a postdoc at UC Berkeley.

Video

Chat is not available.