Reproduce “Practical Federated Gradient Boosting Decision Trees” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Qinbin Li, Zeyi Wen, Bingsheng He
A framework for horizontal federated Gradient Boosting Decision Trees is presented, employing Locality-Sensitive Hashing for privacy-preserving similarity discovery and a Weighted Gradient Boosting method. This approach achieves predictive accuracy on par with models trained on unshared data, while maintaining low computational and communication costs within a relaxed privacy model.
Key insights
- Efficiently discovering and leveraging data similarity through LSH can substantially enhance GBDT accuracy in horizontal federated learning, surpassing local-only models.
- A relaxed privacy model offers a practical pathway to significantly more efficient and deployable federated GBDT solutions, compared to strict cryptographic approaches.
- Weighted Gradient Boosting effectively integrates distributed information from similar instances, facilitating robust collaborative learning without raw data exposure.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.