Research
Lisa is broadly interested in optimizaiton and machine learning for prescriptive applications. Currently, her work involves collaborators in healthcare, including Massachusetts General Hospital (MGH) and Hartford Hospital.
(*) = co-first author
Submitted or in preparation
(alphabetical) D. Bertsimas, L. Everest, V. Stoumpou. Interpretable Prescriptive Neural Networks, submitted to Machine Learning, 2024.
P. Paschalidis, V. Stoumpou, L.Everest, Y. Ma, …, D. Bertsimas. Towards Optimal Valve Prescription for Transcatheter Aortic Valve Replacement (TAVR) Surgery: A Machine Learning Approach, in preparation for Health Care Management Science, 2024
Current
- Optimal Control of Markov Decision Processes: A Machine Learning Approach
- Exploring machine learning approach for the optimal control of general Markov decision processes (MDP)
- Using numerical solutions of MDP problems as a training set for machine learning model to learn the control policies – the idea being while offline training could take a long time, online application could be fast
- Multi-Modal Machine Learning with Hartford HealthCare:
- Developing predictive multimodal models using real-world tabular (demographic, labs, disease history), language (discharge summaries, nurse notes), and images (medical scans)
- Medical specialties
- Orthopedics: Fragility Hip Fractures’ Readmission, Transfusion, Time to Surgery
- Pharmacy: Medication-Related Readmission