It would be beneficial to include a dedicated example in Flower demonstrating Federated Learning (FL) for time series data. Time series is a crucial data type in various domains, such as finance, healthcare, IoT, and energy systems.
This example could be made general or tailored to a specific domain (e.g. by using a specific dataset for example). Example datasets can be found on HuggingFace or UCI.
This would be a great contribution to the Flower framework for anyone who is interested to open a PR. The Flower team will help you throughout the process to successfully merge an example and gain respect from the open-source community.
Hi, I’m using Flower for our time-series forecasting research project and I have a working example with a HuggingFace dataset. What I currently look into is the data partitioning in combination with the slicing of data in temporal windows that is often required for time series algorithms, especially for neural network based ones.
In case you still want to include such an example please let me know.
We’d love to see such implementation and I’d be happy to review the pull request/code example. An initial example would not need to be too extensive and can also come to be updated, thus if you’d like - feel free to initially submit the working code example for review.
Are you familiar with the GitHub PR process? If not, I’m happy to schedule a brief call to go over it.
Thank you for your message and for offering to review the implementation! I’ll prepare an initial working example to submit it as a pull request soon. I appreciate the flexibility to start with something simple and iterate from there.
I’m a Git user but I’m not very familiar with the GitHub PR process, so I’d appreciate some instructions or a brief call to go over it. Please let me know what works best for you.