How would the performance of federated learning (FL) compare to the performance of centralized machine learning (ML), when the data is independent and identically distributed (i.i.d.)?

Moreover, what is the difference in the performance of FL when the data is i.i.d. as compared to non-i.i.d?

  • $\begingroup$ @nbro you are wrong it's not FD! $\endgroup$
    – Jared
    Mar 23 at 19:38
  • $\begingroup$ Ha, sorry, my bad, I didn't realise I had written D instead of L. It was a typo. Thanks for pointing out. Feel free to edit your post even further to clarify it. $\endgroup$
    – nbro
    Mar 23 at 20:15

There are some works that do this comparison. Briefly, it's been observed that the performance of models trained via FL drops as data distributions between participating agents differ. When data is IID-like though, performance is comparable to centralized training. Some works that I'm aware of are as follows:

  1. Overcoming Forgetting in Federated Learning on Non-IID Data
  2. Improving Accuracy of Federated Learning in Non-IID Settings
  3. Federated Learning with Non-IID Data

There are probably many more around. It's an active area of research.

  • $\begingroup$ Thanks for your information! one more question, do you have specific resources for "performance is comparable to centralized training."? $\endgroup$
    – Jared
    Mar 23 at 8:03
  • $\begingroup$ that’s what the results indicate. look at the experiments of paper 2 and 3. $\endgroup$
    – SpiderRico
    Mar 23 at 9:19

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.