Many of today’s most pressing challenges involve complex optimization problems—often NP-hard and extremely difficult to solve at scale. These problems arise in diverse, high-impact domains, including renewable energy management, healthcare resource allocation, and global supply chain logistics. While ML has the potential to transform both applied large-scale optimization and theoretical combinatorial optimization, algorithmic reasoning remains a significant challenge for artificial intelligence.
Our lab’s research is driven by the observation that, in practice, there is often a wealth of data specific to the application domain that can be leveraged to optimize algorithmic performance. For instance, the scheduling problems that an electric grid operator faces will change daily, but not drastically: although demand will vary, the network structure will remain largely stable. Thus, there is likely underlying structure linking the problems that arise within a given application domain. We aim to use ML to uncover and leverage this structure to optimize algorithmic performance.
Applications are due on April 3, 2025 by midnight AoE. Instructions below.