Cases and applications of federated learning

[falibabaei]

Unconference Federated Learning

Title

Cases and applications of federated learning

Description

Centralized Learning (CL) in Machine Learning refers to the traditional approach where all data is gathered and stored in a central location to train a machine learning model. It involves collecting and combining data from multiple sources into a single dataset before training the model. There are some challenges using CL ML such as
Data Flow Management: Manage the transfer of large volumes of diverse data quickly and accurately across different organizations.
Scalability
Communication Overhead
Intense competition within the industry.
Data Privacy: Ensuring compliance with strict data protection regulations, such as the GDPR1 and EU AI ACT2.
Federated Learning (FL) is a privacy-preserving machine learning paradigm introduced to address concerns related to centralized model training. It facilitates model training in a network of decentralized devices and favours data privacy. FL is proven to be beneficial not only for privacy protection but also for scenarios with limited or heterogeneous resources, such as network bandwidth and computational power. This versatility makes FL suitable for various real-world applications, especially in decentralized or resource-constrained environments. As FL continues to evolve, its potential impact on various areas, including healthcare, finance, and smart cities, becomes clearer.

In this unfconference we are welcoming to talk about

• Real-world cases where federated learning is beneficial
• Pros and challenges of Federated learning
• Available frameworks for applying FL
• Potentials for collaboration

We want to discuss the topic of federated learning in an open format with you, find real case scenarios, and connect with interested people.

Organizational

Host(s)

Khadijeh Alibabaei <[email protected]>
Leonhard Duda [email protected]

Format

Introduction talk followed by open discussion, brainstorming, etc.
https://youtu.be/v7blpOB92C8

Timeframe

1-2h

Number of participants

3-20

Material

Beamer, flipchart, markers, post-its

[falibabaei]

[SusanneWenzel]

Dear @falibabaei ,

many thanks for your contribution! We hope your session was successful and constructive!

To make this not only a nice experience during the conference but a sustainable format, we would like to release a short result on the Helmholtz AI Website. Please provide us with a little report about the discussion in your session:

• One or two intro sentences about the topic and the respective format of your session
• What was most controversial, or about what did all participants agree on?
• What were the key ideas?
• Do you see any follow up on this discussion?

Depending on the intensity of the session it can be shorter (1/3 page) or longer (one pager).

Please post your report here by end of next week. Ina already confirmed to release that at a HAICON24 subpage at Helmholtz.ai.

Best regards,
@helenehoffmann and Susanne

[LeoDuda]

Dear @SusanneWenzel,

in the following our thoughts and comments of our unconference experience, that we really enjoyed.

On the unconference with the subject of Cases and applications of federated learning, more than 10 people participated.

• Some slides from the AI4EOSC webinar [1] were shown.
• We started talking about our experience working with NVFlare [2].
• Most of the participants already knew something about Federated Learning or had at least a basic understanding of it.
• There was a discussion about the simplicity of using Federated Learning on a real world example and implementing it into a use case.
• Some participants pointed out that they wanted to use Federated Learning in a use case which includes hospitals, but they experienced the problem that no data scientist was working there, which could administer the technical part for it.
• Other participants mentioned a different issue: the clients did not trust this method, they are too skeptical and don’t want to open connections to other participants, due to this lack of trust.
• The participants also agreed that within a use case where privacy is not handled that strictly, the implementation of Federated Learning is much simpler.
• There were also conversations about AI4EOSC project [3] and the platform that provides FL.
• One of the participants asked if she could use FL for training models other than deep learning models.

Due to time limits the conversation and discussion couldn’t proceed any further and some open questions remained. We also would have liked to gather new ideas of areas where Federated Learning can be introduced. But also some approaches on how to handle that lack of trust that exists within some areas where privacy matters a lot.

[1] Alibabaei, K. (2024, April 22-23). Basics of Federated Learning: Tips and Tricks. AI4EOSC Webinar: Introduction to Federated Learning. Online. Note: 46.21.02; LK 01.
[2] Roth, H. R., Cheng, Y., et. al. (2023). NVIDIA FLARE: Federated Learning from Simulation to Real-World. IEEE Data Eng. Bull., Vol. 46, No. 1. https://doi.org/https://doi.org/10.48550/arXiv.2210.13291
[3] https://ai4eosc.eu/

Best regards,
@falibabaei and Leo