Research map / Federated learning and privacy / Personalized and heterogeneous FL
QuPeD: Quantized Personalization via Distillation with Applications to Federated Learning
Kaan Ozkara, Navjot Singh, Deepesh Data, Suhas Diggavi
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 |
|---|---|---|---|
| update quantization centers instead of fixed centers (1bit) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-20 | test accuracy (%): 90.69 -> 91.17 | Improved |
| update quantization centers instead of fixed centers (1bit) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-20 | test accuracy (%): 87.82 -> 91.17 | Improved |
| update quantization centers (1bit) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-32 | test accuracy (%): 91.55 -> 92.20 | Improved |
| use 2 bits instead of 1 bit (ablation) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-20 | test accuracy (%): 91.17 -> 91.45 | Improved |
| use 2 bits instead of 1 bit (ablation) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-32 | test accuracy (%): 92.20 -> 92.47 | Improved |
| quantize to 2 bits versus full precision (ablation) | centralized quantization Algorithm 1 · CIFAR-10 ResNet-20 | test accuracy (%): 92.05 -> 91.45 | Hurt |
| QuPeD collaborative training versus local training, full precision | QuPeD · FEMNIST, CNN1(FP)+CNN2(FP) | test accuracy (%): 93.41 -> 97.44 | Improved |
| QuPeD versus local training with 2-bit CNN1 and FP CNN2 | QuPeD · FEMNIST | test accuracy (%): 92.70 -> 97.01 | Improved |
| QuPeD versus local training with 2-bit CNN1 and 2-bit CNN2 | QuPeD · FEMNIST | test accuracy (%): 91.11 -> 96.64 | Improved |
| QuPeD versus local training, full precision | QuPeD · CIFAR-10, CNN1(FP)+CNN2(FP) | test accuracy (%): 72.81 -> 75.50 | Improved |
| QuPeD versus local training with 2-bit CNN1 and 2-bit CNN2 | QuPeD · CIFAR-10 | test accuracy (%): 72.15 -> 74.64 | Improved |
| quantize CNN1 to 2 bits in resource heterogeneous setting (ablation) | QuPeD · CIFAR-10 | test accuracy (%): 75.50 -> 75.08 | No clear effect |
Closest papers by meaning
- Fed-QSSL: A Framework for Personalized Federated Learning under Bitwidth and Data Heterogeneity (AAAI 2024 · similarity 0.90)
- Mixed-Precision Quantization for Federated Learning on Resource-Constrained Heterogeneous Devices (CVPR 2024 · similarity 0.88)
- DAdaQuant: Doubly-adaptive quantization for communication-efficient Federated Learning (ICML 2022 · similarity 0.88)
- Decentralized Directed Collaboration for Personalized Federated Learning (CVPR 2024 · similarity 0.86)
- A Bayesian Approach for Personalized Federated Learning in Heterogeneous Settings (NeurIPS 2024 · similarity 0.86)
- Understanding the Statistical Accuracy-Communication Trade-off in Personalized Federated Learning with Minimax Guarantees (ICML 2025 · 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.
