FedSaw: Communication-Efficient Cross-Silo Federated Learning with Adaptive Compression
Bibliographic record
Abstract
Cross-silo federated learning (FL) is an emerging approach for institutions to collaboratively train a machine learning model without sharing their siloed data. However, it conventionally requires institutions to centrally store their clients’ data, posing a threat to clients’ privacy. This paper thus studies a preferable setting to leave data distributed to clients, where traditional FL is used not only across institutions but also among their clients. As this new setting inherits or even exacerbates the major problem of communication inefficiency in the traditional setting, we explore the feasibility of leveraging two compression techniques, pruning and quantization, to improve communication efficiency. Starting by applying off-the-shelf pruning and quantization mechanisms, we observe that they could largely reduce communication overhead with a negligible reduction, sometimes even a slight increase, in training performance. By mathematically analyzing the impact of compression on the performance of the trained model, we find that pruning and quantizing with a proper amount can offset possible performance degradation due to non-i.i.d data. Based on this finding, we propose FedSaw, a new cross-silo FL framework that can improve communication efficiency by adaptively tuning the pruning amount and quantizing updates throughout training. In our extensive evaluation with six benchmark datasets, FedSaw consistently outperformed its state-of-the-art competitors. It decreased the wall-clock training time and communication overhead used for converging to the target accuracy by up to 86.7% and 91.5%, respectively.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.012 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.002 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.004 |
| Open science | 0.004 | 0.004 |
| Research integrity | 0.002 | 0.002 |
| Insufficient payload (model declined to judge) | 0.002 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".