Research map / Federated learning and privacy / Communication-efficient FL
DeepReduce: A Sparse-tensor Communication Framework for Federated Deep Learning
Hang Xu, Kelly Kostopoulou, Aritra Dutta, Xin Li, Alexandros Ntoulas, Panos Kalnis
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 |
|---|---|---|---|
| Add Bloom-filter index compression with policy P0 to Top-r (10%) | LSTM · Stack Overflow | data volume: 0.2033 -> 0.1425 (S2C); 0.2033 -> 0.1426 (C2S) | Improved |
| Add QSGD value compression with Bloom-filter P0 indices | LSTM · Stack Overflow | data volume: 0.2033 -> 0.0621 | Improved |
| Add DeepReduce BF-P0 to Top-r sparsification | ResNet-20 · CIFAR-10 | Top-1 accuracy | No clear effect |
| Use Bloom-filter policy P0 instead of plain Top-r (ablation) | ResNet-20 · CIFAR-10 | training epochs to convergence: 33% fewer data | Improved |
| Replace Bloom-filter policy P0 with naïve Bloom-filter reconstruction (ablation) | ResNet-20 · CIFAR-10 | Top-1 accuracy | Hurt |
| Use Bloom-filter policy P1 rather than P0 (ablation) | Bloom-filter index compression · Sparse gradients | reconstruction error | Hurt |
| Use Bloom-filter policy P2 rather than P1 (ablation) | Bloom-filter index compression · Sparse gradients | reconstruction error | Improved |
| Apply Fit-Poly value compression to Top-r (1%) | ResNet-20 · CIFAR-10 | Top-1 accuracy | No clear effect |
| Replace Fit-Poly with Fit-DExp (ablation) | ResNet-20 · CIFAR-10 | data volume: roughly 50% versus around 40% | Improved |
| Replace Fit-Poly with Fit-DExp (ablation) | ResNet-20 · CIFAR-10 | compression computational overhead: roughly 3.5x more | Hurt |
| Apply DR_QSGD_BF-P0 | LSTM · Stack Overflow | Test Accuracy: 0.1856 -> 0.1836 | No clear effect |
| Apply DR_BF-P0 | NCF · Movielens-20M | Best Hit Rate | No clear effect |
Closest papers by meaning
- Achieving Lossless Gradient Sparsification via Mapping to Alternative Space in Federated Learning (ICML 2024 · similarity 0.88)
- SpaFL: Communication-Efficient Federated Learning With Sparse Models And Low Computational Overhead (NeurIPS 2024 · similarity 0.87)
- LoCoDL: Communication-Efficient Distributed Learning with Local Training and Compression (ICLR 2025 · similarity 0.87)
- Distribution-Aware Tensor Decomposition for Compression of Convolutional Neural Networks (NeurIPS 2025 · similarity 0.86)
- FedRTS: Federated Robust Pruning via Combinatorial Thompson Sampling (NeurIPS 2025 · similarity 0.86)
- Convolutions and More as Einsum: A Tensor Network Perspective with Advances for Second-Order Methods (NeurIPS 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.
