Machine learning is widely used to support decision making in a range of application domains. In safety critical applications, such as in healthcare, it is vital to be able to explain the reasoning of the machine learning system and to guarantee that certain dangerous or unwanted behaviors will not appear. In this research project, the goal is to investigate how mathematical optimization can be used to train interpretable machine learning models that are guaranteed to satisfy a set of required constraints.

The SUDS Scholar working on this project will:

• Read, implement, and empirically evaluate state-of-the-art mathematical optimization models for machine learning from the literature.
• Develop new mathematical models and extend existing models from the literature with an emphasis on interpretability and supporting domain-specific constraints; Efficiently implementing these models using state-of-the-art optimization software.
• Run experiments to evaluate the efficiency and effectiveness of the mathematical models using standard benchmarks and different real-world datasets.