DIStributed COllaborative learning for Data-driven Humanitarian Response in Insecure and Volatile Environments
We are working with the International Committee of the Red Cross (ICRC) to create DISCO: an open-access, open-source DIStributed COllaborative learning platform, allowing multiple parties to collaboratively build machine learning models without sharing any original data.
Data collected in humanitarian responses is rarely shared in real-time and is often never shared at all for a range of well-considered reasons, such as privacy, ethicolegal constraints, limited bandwidth, and security concerns. Yet this data could be extremely valuable to inform humanitarian action, evaluate its impact, and improve transparency and accountability. This project will adapt DISCO to meet the unique constraints of humanitarian settings to enable continuous learning for data-driven humanitarian response. Co-designed with our partner lab in Switzerland @ EPFL computer science school (Prof Martin Jaggi, Machine Learning and Optimization lab)
