Neural Set Function Extensions: Learning with Discrete Functions in High Dimensions

Explore a comprehensive lecture on Neural Set Function Extensions, focusing on learning with discrete functions in high dimensions. Delve into the challenges of integrating discrete domain functions into neural networks and discover innovative solutions for set functions. Examine a framework for extending set functions onto continuous domains and learn how to convert low-dimensional extensions into high-dimensional representations. Gain insights into unsupervised neural combinatorial optimization and its applications. Follow along as the speaker covers motivation, existing techniques, extension examples, lifting to higher dimensions, neural extensions, and real-world applications. Conclude with a discussion on future work, open problems, and a Q&A session to deepen your understanding of this cutting-edge research in machine learning and combinatorial optimization.

- Intro
- Outline
- Motivation
- Continuous vs. Discrete: Set Functions and Potential Solutions
- Overview of Existing Techniques
- Extensions: Main Idea and Intuition
- Examples of Extensions
- Lifting to Higher Dimensions
- Neural Extensions
- Background: Low and High Dimensional Extensions
- Applications of Extensions and Experiments
- Future Work and Open Problems
- Q+A