Introduction

Welcome to the Mila GFlowNet Workshop!

During this workshop, we will explore the fascinating world of generative flow networks (GFlowNets). GFlowNets are trained to take sequential actions to sample from a distribution specified by an energy function or reward function, allowing us to generate complex objects through a sequence of steps. GFlowNets have been applied to a variety of domains, including molecule and protein design, causal discovery, and combinational optimization, as well as to augment the capabilities of generative models for vision and language.

GFlowNets are a powerful framework related to concepts from Monte-Carlo Markov chain methods, reinforcement learning, generative models, and amortized variational methods.

Throughout the workshop, we will discuss the backgrounds, theoretical foundations, techniques, and applications of GFlowNets, providing you with a comprehensive understanding of this exciting new field. We hope you enjoy the workshop and find inspiration in the potential of Generative Flow Networks.

Goal

The goal of the workshop is to establish an inclusive atmosphere where researchers and practitioners from diverse backgrounds can come together to have a better understanding and deeper insights about the theory, methods, and applications of GFlowNets.

The workshop will delve into several topics, such as pre-requisites, theoretical backgrounds, algorithm designs, and practical applications. Additionally, the workshop will scrutinize the limitations of present methods, benchmarks, and assumptions, and explore ways to enhance existing methods to facilitate wider applications in fields like molecular discovery, biological sequence design, and other related areas.

Attendance

• When: The Mila GFlowNet Workshop consists of a three-full-day conference at Mila from November 8-10, 2023.
  • Day 1: (optional) refresher on statistics, probabilistic and generative modeling
  • Day 2: An introduction to the theory of GFlowNets
  • Day 3: An overview of existing applications

• Where:

  • In-person participation: The workshop will be an in-person event and to be held at the Agora of Mila, located at 6650 rue Saint-Urbain, Montreal, QC, H2S 3H1.
  • Online perticipation: You can also join the workshop virtually at [TBD].

Contact

If you have questions or feedback, please send an email to gflownet.workshop@mila.quebec.