Transferrable and Interpretable ML for Healthcare

Research Funding:

I and my academic collaborators, Douglas A. Lauffenburger (MIT) and Sara Magliacane (University of Amsterdam), were awarded an exploratory research grant (150K USD) to investigate new challenges in Cross-Species Translation of COVID-19 Systems Serology Data for Infection and Vaccine Treatment by the MIT-IBM Watson AI Lab.

Idea and Motivation:

I am interested in application of ML to healthcare. Two important desiderata of ML models for healthcare applications are (1) transferability where knowledge learned from one domain can be transferred effectively to another and (2) interpretability via data prototyping where models whose predictions at prototypical data points are interpretable.

Relevant papers (in collaboration with Danica Xiao and Jimeng Sun:

AID: Active Distillation Machine to Leverage Pre-Trained Black-Box Models in Private Data Settings (WWW-21) Paper

CHEER: Rich Model Helps Poor Model via Knowledge Infusion (TKDE-20) Paper

CASTER: Predicting Drug Interactions with Chemical Substructure Representation (AAAI-20) Paper MIT Tech Review Tech Republic

Older works in relevant topics:

RDPD: Rich Data Helps Poor Data via Imitation (IJCAI-19) Paper

DDL: Deep Dictionary Learning for Predictive Prototyping (IJCAI-19) Paper