Research Projects and Directions

Decision-focused Learning - Integrating Domain Knowledge to Machine Learning Pipelines

Machine learning models have achieved significant success in many industrial applications and social challenges, including natural language processing, computer vision, time series analysis, and recommendation systems. To adapt to different applications, incorporating structural information and domain knowledge in applications into machine learning models is an important element of the training process. But it often relies on fine-tuning and feature-engineering without a systematic approach to adapt to various applications. On the other hand, operational research is an application-driven approach, where optimization problems are formulated based on the knowledge and constraints of targeted applications to derive actionable solutions. Optimization formulations can capture structural information and domain knowledge in applications, but the non-differentiability and the complex operational processes in optimization make it hard to integrate into machine learning models.

Online Learning

Online learning is a different way to adopt to domain knowledge. It is always important to continuously improve the information received from interacting with the environment. I focus on applying online learning to optimization problems induced from different social challenges. Different optimization formulations produce different unique challenges in learning the unknown information, including different information received from interaction and different constraints on interacting with the environment. Properly handling these unique operational challenges is crucial for successfully shifting online learning solutions to social impact.

Game Theory and EconCS

Game theory is a powerful tool for modeling strategic agent behavior. It is particularly important to consider agent interaction in a large multi-agent system, e.g., wildlife conservation, security games, and auction design, which are often modeled as Stackelberg games with various number of agents. Stackelberg games can often be formulated as an optimization problem but with expensive computation cost to solve it efficiently. I focus on improving the scalability of the existing Stackelberg game solution by providing approximate algorithms or using novel technique from differentiable optimization to provide a gradient-based solution to solve the optimization problem.