186 points by ml_researcher 6 months ago flag hide 15 comments
john_doe 6 months ago next
Great work! This could really change the way we approach machine learning. I'm looking forward to seeing more about this framework and its use cases.
sentdex 6 months ago next
Don't forget about potential roadblocks related to data privacy, community management, and optimization. Looking forward to hearing more about solutions to these challenges.
buddhidev 6 months ago next
Most of the time, it's recommended to use existing tools like Tensorflow Privacy and Flower. They already handle many of the privacy/optimization challenges.
opensource_dave 6 months ago prev next
OpenMined is also working in this space. Have you considered combining forces?
buddhidev 6 months ago next
We've actually spoken with the OpenMined team, but are still discussing ways to best collaborate while maintaining each project's unique approach.
fancy_federator 6 months ago prev next
We've also tried Flower, but found it a bit difficult to match our team's requirements. This led us to build our own framework that's easier to use.
riskyclick 6 months ago prev next
I remember reading about federated learning in the latest Deep Learning Book. How does this framework differ from others?
machine_mendel 6 months ago next
Our framework mainly focuses on scalability and usability, making it easy for teams to work together and deploy federated solutions faster.
quantum_dude 6 months ago prev next
I briefly looked at Tensorflow Privacy, and it appears quite complex for a practitioner. How does this framework simplify these privacy concepts?
harry_potter 6 months ago prev next
Fascinating! I've been working on a similar project, and I think federated learning will become a big part of the future. Thanks for sharing!
crazy_coder 6 months ago prev next
I'm curious if you've seen the latest paper from the Tensorflow team. Should we expect integration with their libraries?
the_real_facebook 6 months ago next
Suppose we consider the possibility of integrating with Tensorflow or PyTorch in the future. But our initial focus is on building a strong open-source user base.
data_witch 6 months ago prev next
Awesome post! Any rough estimates on when we might expect functional documentation or examples?
ai_geek 6 months ago next
We're looking forward to sharing more over the next few weeks. Follow our Twitter account for updates!
generative_gal 6 months ago prev next
Federated learning will be our workhorse for training models without relying on client-side data storage for years to come!