Consolidate DISCO fundamentals and implement new features.
Prepare DISCO for the Global Digital Health Forum in Dec 2024.

DISCO should be ready for the following uses (from the GDHF abstract):

1. Understanding Current Workflows in Collaborative Learning
   Attendees will learn how DISCO leverages the fundamental principles of federated learning (FL) to conduct private training. In particular, they will understand how classical FL preserves privacy compared to conventional machine learning, and what additional measures we take to ensure further privacy guarantees. They will also learn how heterogeneous datasets can affect model quality and what steps can be taken to reduce bias.

2. Mastering the Use of the DISCO Platform
   Participants will learn how to create and deploy a machine learning task (a DISCOllaborative) on the DISCO platform. They will learn how to create a task, upload training and test data, and recruit other peers to their DISCOllaborative. Participants will understand the difference between collaborative learning that uses a central server to coordinate model updates (centralized FL) versus peer-to-peer learning (decentralized FL), both workflows of which are available in DISCO. We will review how to set various hyperparameters for model training that will influence model convergence and affect privacy guarantees.

3. Transforming privacy-preserving AI for healthcare
   Participants will explore how DISCO can improve data privacy, improve technological equity and accessibility, and build powerful AI models using the knowledge of many decentralized datasets.