Federated Visual Classification with Real-World Data Distribution

  title={Federated Visual Classification with Real-World Data Distribution},
  author={Tzu-Ming Harry Hsu and Qi and Matthew Brown},
Federated Learning enables visual models to be trained on-device, bringing advantages for user privacy (data need never leave the device), but challenges in terms of data diversity and quality. Whilst typical models in the datacenter are trained using data that are independent and identically distributed (IID), data at source are typically far from IID. Furthermore, differing quantities of data are typically available at each device (imbalance). In this work, we characterize the effect these… 

Improving Generalization in Federated Learning by Seeking Flat Minima

This work investigates behavior through the lens of geometry of the loss and Hessian eigenspectrum, linking the model’s lack of generalization capacity to the sharpness of the solution and shows that training clients locally with Sharpness-Aware Minimization or its adaptive version can sub-stantially improve generalization in Federated Learning and help bridging the gap with centralized models.

Learning to Generate Image Embeddings with User-level Differential Privacy

DP-FedEmb, a variant of federated learning algorithms with per-user sensitivity control and noise addition, is proposed to train image embedding models using supervised training data centralized in the datacenter to achieve user-level DP for large image-to-embedding feature extractors.

On the Importance and Applicability of Pre-Training for Federated Learning

It is found that pre-training enables the learned global models under different clients’ data conditions to converge to the same loss basin, and makes global aggregation in FL more stable.

FedNorm: Modality-Based Normalization in Federated Learning for Multi-Modal Liver Segmentation

FedNorm and its extension FedNorm+ , two Federated Learning algorithms that use a modality-based normalization technique, are proposed that show an overall acceptable performance, achieve Dice per patient scores up to 0.961, consistently outperform locally trained models, and are on par or slightly better than centralized models.

On Bridging Generic and Personalized Federated Learning for Image Classification

This paper proposes a novel federated learning framework that explicitly decouples a model’s dual duties with two prediction tasks, and introduces a family of losses that are robust to non-identical class distributions, enabling clients to train a generic predictor with a consistent objective across them.

Dubhe: Towards Data Unbiasedness with Homomorphic Encryption in Federated Learning Client Selection

This work proposes a pluggable system-level client selection method named Dubhe, which allows clients to proactively participate in training, meanwhile preserving their privacy with the assistance of additively homomorphic encryption, and shows that Dubhe is comparable with the optimal greedy method on the classification accuracy.

On Bridging Generic and Personalized Federated Learning

This paper proposes a novel federated learning framework that explicitly decouples a model’s dual duties with two prediction tasks, and introduces a family of losses that are robust to non-identical class distributions, enabling clients to train a generic predictor with a consistent objective across them.

No Fear of Heterogeneity: Classifier Calibration for Federated Learning with Non-IID Data

A novel and simple algorithm called Classifier Calibration with Virtual Representations (CCVR), which adjusts the classifier using virtual representations sampled from an approximated gaussian mixture model, which achieves state-of-the-art performance on popular federated learning benchmarks including CIFAR-10, CIFar-100, and CINIC-10.

Federated Learning with Privacy-Preserving Ensemble Attention Distillation

This work proposes a privacy-preserving FL framework leveraging unlabeled public data for one-way offline knowledge distillation in this work and achieves very competitive performance with more robust privacy preservation based on extensive experiments on image classification, segmentation, and reconstruction tasks.

On Pre-Training for Federated Learning

It is found that pretraining does largely close the gap between F ED A VG and centralized learning under non-IID data, but this does not come from alleviating the well-known model drifting problem in FED A VG ’s local training.



Google Landmarks Dataset v2 – A Large-Scale Benchmark for Instance-Level Recognition and Retrieval

This work introduces the Google Landmarks Dataset v2 (GLDv2), a new benchmark for large-scale, fine-grained instance recognition and image retrieval in the domain of human-made and natural landmarks, and demonstrates the suitability of the dataset for transfer learning by showing that image embeddings trained on it achieve competitive retrieval performance on independent datasets.

Communication-Efficient Learning of Deep Networks from Decentralized Data

This work presents a practical method for the federated learning of deep networks based on iterative model averaging, and conducts an extensive empirical evaluation, considering five different model architectures and four datasets.

Measuring the Effects of Non-Identical Data Distribution for Federated Visual Classification

This work proposes a way to synthesize datasets with a continuous range of identicalness and provide performance measures for the Federated Averaging algorithm, and shows that performance degrades as distributions differ more, and proposes a mitigation strategy via server momentum.

SCAFFOLD: Stochastic Controlled Averaging for Federated Learning

This work obtains tight convergence rates for FedAvg and proves that it suffers from `client-drift' when the data is heterogeneous (non-iid), resulting in unstable and slow convergence, and proposes a new algorithm (SCAFFOLD) which uses control variates (variance reduction) to correct for the ` client-drifts' in its local updates.

Advances and Open Problems in Federated Learning

Motivated by the explosive growth in FL research, this paper discusses recent advances and presents an extensive collection of open problems and challenges.

SCAFFOLD: Stochastic Controlled Averaging for On-Device Federated Learning

A new Stochastic Controlled Averaging algorithm (SCAFFOLD) which uses control variates to reduce the drift between different clients and it is proved that the algorithm requires significantly fewer rounds of communication and benefits from favorable convergence guarantees.

Real-World Image Datasets for Federated Learning

This paper introduced a real-world image dataset and implemented two mainstream object detection algorithms (YOLO and Faster R-CNN) and provided an extensive benchmark on model performance, efficiency, and communication in a federated learning setting.

The Non-IID Data Quagmire of Decentralized Machine Learning

SkewScout is presented, a system-level approach that adapts the communication frequency of decentralized learning algorithms to the (skew-induced) accuracy loss between data partitions and it is shown that group normalization can recover much of the accuracy loss of batch normalization.

Federated Learning: Challenges, Methods, and Future Directions

The unique characteristics and challenges of federated learning are discussed, a broad overview of current approaches are provided, and several directions of future work that are relevant to a wide range of research communities are outlined.

Large-scale Landmark Retrieval/Recognition under a Noisy and Diverse Dataset

This work presents a novel landmark retrieval/recognition system, robust to a noisy and diverse dataset, based on deep convolutional neural networks with metric learning, trained by cosine-softmax based losses.