News from healthyml.org

Views Can Be Deceiving: Improved SSL Through Feature Space Augmentation | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

When Style Breaks Safety: Defending LLMs Against Superficial Style Alignment | Healthy ML

Yuxin Xiao is a Ph.D. candidate at MIT IDSS. His research focuses on building safe, robust, and trustworthy LLMs and advancing their reasoning and decision-making capabilities for healthcare and other high-stakes applications. Yuxin obtaine…

healthyml.org·

WRING Out The Bias: A Rotation-Based Alternative To Projection Debiasing | Healthy ML

Cassandra is a third year Ph.D. student in the Medical Engineering and Medical Physics (MEMP) program at Harvard and MIT. Cassandra’s research focuses on equitable machine learning models that target underserved health conditions. She is pa…

healthyml.org·

Opportunistic screening of type 2 diabetes with deep metric learning using electronic health records | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

On Group Sufficiency Under Label Bias | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

LEMoN: Label Error Detection using Multimodal Neighbors | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

Self-supervised contrastive learning of protein representations by mutual information maximization | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

Deep Metric Learning for the Hemodynamics Inference with Electrocardiogram Signals | Healthy ML

Hyewon Jeong is a Ph.D. student in EECS at MIT. Her primary research focus has been on applying machine learning models to solve real-world clinical problems, specifically tasks from time-series EHR data, signal data to multi-modal data. Sh…

healthyml.org·

A comparison of approaches to improve worst-case predictive model performance over patient subpopulations | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

Change is Hard: A Closer Look at Subpopulation Shift | Healthy ML

Haoran is a fifth year PhD student in EECS at MIT. He is generally interested in building robust machine learning models that maintain their performance and fairness across out-of-distribution environments, as well as applying such models t…

healthyml.org·

In the Name of Fairness: Assessing the Bias in Clinical Record De-identification | Healthy ML

Yuxin Xiao is a Ph.D. candidate at MIT IDSS. His research focuses on building safe, robust, and trustworthy LLMs and advancing their reasoning and decision-making capabilities for healthcare and other high-stakes applications. Yuxin obtaine…

healthyml.org·

Creating actionable insights in human health. | Healthy ML

We work on robust machine learning model that can efficiently and accurately model events from healthcare data, and investigate best practices for multi-source integration, and learning domain appropriate representations. The labels we obta…