Skip to main content

Datasets and Models

Industrial Support​

FEDML Provides Cross-silo FL Dataset with Okwin, Allowing Research to Face Real Scenario

  • Blog Introduction

https://medium.com/@FedML/fedml-provides-cross-silo-fl-dataset-with-okwin-allowing-research-to-face-real-scenario-e64b57c5d767

  • Publication in NeurIPS

Publication: https://openreview.net/forum?id=GgM5DiAb6A2

  • Source Code

https://github.com/FedML-AI/FedML/tree/master/python/examples/federate/prebuilt_jobs/healthcare

FEDML Provides Cross-silo FL Dataset with Okwin, Allowing Research to Face Real Scenario

Research Support​

FedML supports comprehensive research-oriented (synthetic and public) FL datasets and models, including four representative synthetic FL datasets used by top-tier publications:

  • EMNIST: EMNIST dataset extends MNIST dataset with upper and lower case English characters.

  • CIFAR-100: CIFAR-100 dataset consists of 100 image classes with each containing 600 images.

  • Shakespeare: Shakespeare dataset is built from the collective works of William Shakespeare.

  • Stack Overflow: Stack Overflow dataset originally hosted by Kaggle consists of questions and answers from the website Stack Overflow. This dataset is used to perform two tasks: tag prediction via logistic regression and next word prediction.

Datasets with downloading service and API provided​

CV​

  • MNIST

  • cifar10

  • cifar100

  • fed_cifar100

  • fed_emnist

  • cinic10

  • ImageNet

  • Landmarks

NLP​

  • shakespeare

  • fed_shakespeare

  • stackoverflow

Finance​

  • lending_club_loan

  • NUS_WIDE

Other​

  • UCI

  • Synthetic

  • edge_case_examples (tailored for paper "Attack of the Tails: Yes, You Really Can Backdoor Federated Learning")

For a comprehensive dataset list, please check the following APIs:

fedml.data.load(args) (https://github.com/FedML-AI/FedML/tree/master/python/fedml/data) and fedml.model.create(args) (https://github.com/FedML-AI/FedML/tree/master/python/fedml/data)

Their usage in different algorithms are as follows:

Horizontal Federated Learning:​

  • Computer Vision: Federated EMNIST + CNN (2 conv layers)
  • Computer Vision: CIFAR100 + ResNet18 (Group Normalization)
  • Natural Language Processing: shakespeare + RNN (bi-LSTM)
  • Natural Language Processing: stackoverflow (NWP) + RNN (bi-LSTM)
  • Computer Vision: CIFAR10, CIFAR100, CINIC10 + ResNet
  • Computer Vision: CIFAR10, CIFAR100, CINIC10 + MobileNet
  • Computer Vision (linear model): MNIST + Logistic Regression
  • Computer Vision (linear model): Synthetic + Logistic Regression

Vertical Federated Learning:​

  • lending_club_loan + VFL
  • NUS_WIDE + VFL

FedNAS​

  • cross-silo CV: CIFAR10, CIFAR100, CINIC10 + ResNet
  • cross-silo CV: CIFAR10, CIFAR100, CINIC10 + MobileNet

Split Learning:​

  • cross-silo CV: CIFAR10, CIFAR100, CINIC10 + ResNet
  • cross-silo CV: CIFAR10, CIFAR100, CINIC10 + MobileNet