Flower Baselines Documentation

Welcome to Flower Baselines’ documentation. Flower is a friendly federated AI framework.

Join the Flower Community

The Flower Community is growing quickly - we’re a friendly group of researchers, engineers, students, professionals, academics, and other enthusiasts.

Join us on Slack

Flower Baselines

Flower Baselines are a collection of organised directories used to reproduce results from well-known publications or benchmarks. You can check which baselines already exist and/or contribute your own baseline.

Method	Dataset	Tags
dasha	cifar10, mushrooms, libsvm	compression, heterogeneous setting, variance reduction, image classification
depthfl	CIFAR-100	image classification, system heterogeneity, cross-device, knowledge distillation
fedavgm	CIFAR-10, Fashion-MNIST	non-iid, image classification
fedbn	MNIST, MNIST-M, SVHN, USPS, SynthDigits	data heterogeneity, feature shift, cross-silo
feddebug	cifar10, mnist	malicious client, debugging, fault localization, image classification, data poisoning
fedmeta	FEMNIST, SHAKESPEARE	meta learning, maml, meta-sgd, personalization
fedmlb	CIFAR-100, Tiny-ImageNet	data heterogeneity, knowledge distillation, image classification
fednova	CIFAR-10	normalized averaging, heterogeneous optimization, image classification
fedpara	CIFAR-10, CIFAR-100, MNIST	image classification, personalization, low-rank training, tensor decomposition
fedper	CIFAR-10, FLICKR-AES	system heterogeneity, image classification, personalization, horizontal data partition
fedpft	CIFAR-100, Caltech101	foundation-models, pre-trained, one-shot, one-round
fedprox	MNIST	image classification, cross-device, stragglers
fedrep	CIFAR-10, CIFAR-100	image classification, label heterogeneity, personalized federated learning
fedstar	Ambient Context, Speech Commands	Audio Classification, Semi Supervised learning
fedvssl	UCF-101, Kinectics-400	action recognition, cross-device, ssl, video, videossl
fedwav2vec2	TED-LIUM 3	speech, asr, cross-device
fjord	“CIFAR-10”	“Federated Learning”, “Heterogeneity”, “Efficient DNNs”, “Distributed Systems”
flanders	MNIST, FashionMNIST	robustness, model poisoning, anomaly detection, autoregressive model, regression, classification
heterofl	MNIST, CIFAR-10	system heterogeneity, image classification
hfedxgboost	a9a, cod-rna, ijcnn1, space_ga, cpusmall, YearPredictionMSD	cross-silo, tree-based, XGBoost, Classification, Regression, Tabular
moon	CIFAR-10, CIFAR-100	data heterogeneity, image classification, cross-silo, constrastive-learning
niid_bench	CIFAR-10, MNIST, Fashion-MNIST	data heterogeneity, image classification, benchmark
tamuna	MNIST	local training, communication compression, partial participation, variance reduction

Tutorials

A learning-oriented series of tutorials, the best place to start.

Note

Coming soon

How-to guides

Problem-oriented how-to guides show step-by-step how to achieve a specific goal.

How-to Guides

• Use Baselines
• Contribute Baselines

Explanations

Understanding-oriented concept guides explain and discuss key topics and underlying ideas behind Flower and collaborative AI.

Note

Coming soon

References

Information-oriented API reference and other reference material.

References

• DASHA: Distributed Nonconvex Optimization with Communication Compression and Optimal Oracle Complexity
• DepthFL: Depthwise Federated Learning for Heterogeneous Clients
• Measuring the effects of non-identical data distribution for federated visual classification
• FedBN: Federated Learning on Non-IID Features via Local Batch Normalization
• FedDebug: Systematic Debugging for Federated Learning Applications
• FedMeta: Federated Meta-Learning with Fast Convergence and Efficient Communication
• FedMLB: Multi-Level Branched Regularization for Federated Learning
• Tackling the Objective Inconsistency Problem in Heterogeneous Federated Optimization
• FedPara: Low-rank Hadamard Product for Communication-Efficient Federated Learning
• Federated Learning with Personalization Layers
• FedPFT: One-shot Federated Learning with Foundation Models
• FedProx: Federated Optimization in Heterogeneous Networks
• Exploiting Shared Representations for Personalized Federated Learning
• FedStar: Federated Self-training for Semi-supervised Audio Recognition
• Federated Self-supervised Learning for Video Understanding
• Federated Learning for ASR Based on wav2vec 2.0
• FjORD: Fair and Accurate Federated Learning under heterogeneous targets with Ordered Dropout
• FLANDERS: Protecting Federated Learning from Extreme Model Poisoning Attacks via Multidimensional Time Series Anomaly Detection
• HeteroFL: Computation And Communication Efficient Federated Learning For Heterogeneous Clients
• Gradient-less Federated Gradient Boosting Trees with Learnable Learning Rates
• Model-Contrastive Federated Learning
• Federated Learning on Non-IID Data Silos: An Experimental Study
• TAMUNA: Doubly Accelerated Federated Learning with Local Training, Compression, and Partial Participation