Federated Learning for Time Series Forecasting with Latent Neural ODEs

Nextflow for Flower

Overview

FL4TSF is a dual-purpose repository that brings together:

1. A Nextflow + Flower framework for the systematic exploration of federated learning (FL) in reproducible, configurable workflows.

• Test several federated learning configuration settings and systematically explore the effects on final local, federated, and centralized performance

2. A testbed for Latent Neural ODEs (Neural Ordinary Differential Equations) applied to time-series forecasting for healthcare

• For example, we use it to explore novel custom aggregation functions specific to neural ODEs

How to use it:

main pipeline:

main.nf

nextflow run main.nf -profile <YOUR_PROFILE>,singularity