Systems Lunch: Nico Christianson (California Institute of Technology), End-to-End Learning for Reliable Decision-Making in Energy and Sustainability

Content

Speaker: Nico Christianson (Caltech)

Abstract: Machine learning can significantly improve performance for decision-making under uncertainty in a wide range of domains. However, ensuring robust, risk-aware decisions - a critical need in energy and sustainability applications - requires well-calibrated uncertainty estimates, which can be difficult to achieve in high-capacity prediction models such as deep neural networks. Moreover, in high-dimensional settings, there may be many valid uncertainty estimates, each with their own performance profile - i.e., not all uncertainty is equally valuable for downstream decision-making. To address this problem, we develop an end-to-end framework to learn the uncertainty sets in conditional robust optimization, with robustness and calibration guarantees provided by conformal prediction. We further show that a similar toolkit can enable learning reliable classifiers for convex-structured classification with provable control over false negative rate. We evaluate these end-to-end methodologies on two applications: (1) grid-scale battery storage optimization, and (2) contingency screening in power grids. In both cases, the end-to-end approach enables significant, data-driven performance improvements that improve on conventional learning approaches while maintaining provable robustness and reliability guarantees.

Bio: Nico Christianson is a fifth (and final)-year PhD candidate in Computing and Mathematical Sciences at Caltech, supported by an NSF Graduate Research Fellowship and a PIMCO Data Science Fellowship. His work is broadly focused on decision-making under uncertainty, with a specific focus on developing new algorithms to enable deploying AI and machine learning tools to real-world energy and sustainability problems while ensuring provable guarantees on reliability, robustness, and safety. At Caltech, he is advised by Adam Wierman and Steven Low, and he has interned at Microsoft Research Redmond and collaborated with industry partners, including Beyond Limits and Amazon. Before Caltech, he received his AB in Applied Mathematics at Harvard College.