Multi-Source Adversarial Transfer Learning in Glucose Prediction for Type-2 Diabetic Patients

This repository contains the source code of the "Adversarial Multi-Source Transfer Learning inHealthcare: Application to Glucose Prediction forDiabetic People" paper, by M. DE BOIS et al.

In particular, the PyTorch implementation of a FCN implementing the multi-source adversarial training methodology and the source-model weights can be found in:

• FCDAN: processing/models/fcn.py
• weights: processing/models/weights/ (to keep the repository small, only the most usefull weights have been uploaded)