GithubHelp home page GithubHelp logo

FedMLB about flower HOT 6 OPEN

jafermarq avatar jafermarq commented on June 15, 2024
FedMLB

from flower.

Comments (6)

alessiomora avatar alessiomora commented on June 15, 2024 1

Hi @jafermarq. I confirm I am based in a territory among the ones listed above (Italy).

from flower.

alessiomora avatar alessiomora commented on June 15, 2024

Hi. I would like to provide the implementation of this paper. I would like to firstly focus on reproducing the results for CIFAR-100 (with distribution-based label skew, Dirichlet concentration param $\alpha \in[0.3, 0.6]$) with ResNet-18 as model architecture.

from flower.

jafermarq avatar jafermarq commented on June 15, 2024

Hey @alessiomora , great to hear you want to reproduce FedMLB.
Could you point us to the specific table/plot for those experiments you mention? is this for the 100 client or 500client setup? also, what point of reference would you include for FedMLB? (i.e. what baseline in the FedMLB paper) -- we have some info about this in the first point in our FAQ: https://flower.dev/summer/#faq

from flower.

alessiomora avatar alessiomora commented on June 15, 2024

Hi @jafermarq. I will start by reproducing the left part of Table 1 in the paper, that is the results for CIFAR-100 with (a) Moderate-scale with Dir(0.3), 100 clients, 5% participation and (b) Large-scale with Dir(0.3), 500 clients, 2% participation, for FedAvg and FedMLB. Those results are also plotted in Fig. 2.

Then, I will reproduce the left part of Table 10 (appendix) in the paper, that is the results for CIFAR-100 with (a) Moderate-scale with Dir(0.6), 100 clients, 5% participation and (b) Large-scale with Dir(0.6), 500 clients, 2% participation, for FedAvg and FedMLB. Those results are also plotted in Fig. 6 (c) and 6 (d), in appendix.

I can also reproduce Table 2 (for FedAvg, FedMLB and FedAvg+KD). Those results are also plotted in Fig. 3.

Then, I am planning to reproduce the results with Tiny-Imagenet as well (the right part of Table 1). I will provide details about this later on.

from flower.

jafermarq avatar jafermarq commented on June 15, 2024

Hi @alessiomora This sounds perfect! I've ticked some of the check list above in the issue. Please take a look at the info regarding the point in step 2 about reward eligibility. If all looks good, feel free to it.

from flower.

jafermarq avatar jafermarq commented on June 15, 2024

Many thanks for confirming! Everything in step 1 and 2 is completed. Looking forward to see FedMLB in action 🙌.

from flower.

Related Issues (20)

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.