Research map / Federated learning and privacy / Personalized and heterogeneous FL
CLoVE: Personalized Federated Learning through Clustering of Loss Vector Embeddings
Randeep Bhatia, Nikos Papadis, Murali Kodialam, T. V. Lakshman, Sayak Chakrabarty
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| replace IFCA's lowest-loss assignment with clustering of loss vectors | IFCA · CIFAR-10, label skew 1 | accuracy: 85.7 -> 90.5 | Improved |
| replace IFCA's lowest-loss assignment with clustering of loss vectors | IFCA · MNIST, concept shift | accuracy: 80.0 -> 97.6 | Improved |
| replace IFCA's lowest-loss assignment with clustering of loss vectors | IFCA · CIFAR-10, concept shift | accuracy: 52.7 -> 58.4 | Improved |
| replace IFCA's lowest-loss assignment with clustering of loss vectors | IFCA · MNIST, label skew 1 | ARI: 0.83 -> 1.00 | Improved |
| use loss-vector embeddings instead of FlexCFL's one-shot gradient clustering | FlexCFL · CIFAR-10, concept shift | ARI: 0.22 -> 1.00 | Improved |
| cluster clients by loss vectors instead of using a single global FedAvg model | FedAvg · CIFAR-10, feature skew | accuracy: 52.4 -> 57.1 | Improved |
| cluster clients instead of training a single global FedAvg model | FedAvg · Amazon Reviews | accuracy: 99.0 -> 86.8 | Hurt |
| use CLoVE instead of FlexCFL | FlexCFL · AG News, label skew 2 | accuracy: 60.2 -> 55.1 | Hurt |
| use CLoVE instead of IFCA for autoencoder training | IFCA · MNIST (unsupervised) | reconstruction test loss: 0.024 -> 0.014 | Improved |
| use CLoVE instead of IFCA for autoencoder training | IFCA · CIFAR-10 (unsupervised) | reconstruction test loss: 0.019 -> 0.021 | Hurt |
| use CLoVE's loss-vector clustering instead of CFL-S gradient clustering for fast cluster recovery | CFL-S · MNIST, concept shift | first round with ARI >= 0.9: 60.0 -> 2.0 | Improved |
| vary model initialization (same vs different) and first-round assignment (random vs evaluation-based) (ablation) | CLoVE and IFCA · MNIST unsupervised, 10 clusters | clustering accuracy over time | Improved |
Closest papers by meaning
- An Efficient Framework for Clustered Federated Learning (NeurIPS 2020 · similarity 0.88)
- Enhancing Clustered Federated Learning: Integration of Strategies and Improved Methodologies (ICLR 2025 · similarity 0.88)
- FLUX: Efficient Descriptor-Driven Clustered Federated Learning under Arbitrary Distribution Shifts (NeurIPS 2025 · similarity 0.88)
- Clustered Federated Learning via Gradient-based Partitioning (ICML 2024 · similarity 0.87)
- PeFLL: Personalized Federated Learning by Learning to Learn (ICLR 2024 · similarity 0.87)
- TurboSVM-FL: Boosting Federated Learning through SVM Aggregation for Lazy Clients (AAAI 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
