Adaptive Histogram-Based Gradient Boosted Trees for Federated Learning

Federated Learning (FL) is an approach to collaboratively train a model across multiple parties without sharing data between parties or an aggregator. It is used both in the consumer domain to protect personal data as well as in enterprise settings, where dealing with data domicile regulation and the pragmatics of data silos are the main drivers. While gradient boosted tree implementations such as XGBoost have been very successful for many use cases, its federated learning adaptations tend to be very slow due to using cryptographic and privacy methods and have not experienced widespread use. We propose the Party-Adaptive XGBoost (PAX) for federated learning, a novel implementation of gradient boosting which utilizes a party adaptive histogram aggregation method, without the need for data encryption. It constructs a surrogate representation of the data distribution for finding splits of the decision tree. Our experimental results demonstrate strong model performance, especially on non-IID distributions, and significantly faster training run-time across different data sets than existing federated implementations. This approach makes the use of gradient boosted trees practical in enterprise federated learning.

[1]  Bingsheng He,et al.  Practical Federated Gradient Boosting Decision Trees , 2019, AAAI.

[2]  Tianqi Chen,et al.  XGBoost: A Scalable Tree Boosting System , 2016, KDD.

[3]  Blaise Agüera y Arcas,et al.  Communication-Efficient Learning of Deep Networks from Decentralized Data , 2016, AISTATS.

[4]  Xiang Li,et al.  On the Convergence of FedAvg on Non-IID Data , 2019, ICLR.

[5]  Heiko Ludwig,et al.  IBM Federated Learning: an Enterprise Framework White Paper V0.1 , 2020, ArXiv.

[6]  Thomas Keck,et al.  FastBDT: A Speed-Optimized Multivariate Classification Algorithm for the Belle II Experiment , 2017, Computing and Software for Big Science.

[7]  Tie-Yan Liu,et al.  LightGBM: A Highly Efficient Gradient Boosting Decision Tree , 2017, NIPS.

[8]  Peter Richtárik,et al.  Federated Learning: Strategies for Improving Communication Efficiency , 2016, ArXiv.

[9]  Jianfeng Ma,et al.  Cloud-based Federated Boosting for Mobile Crowdsensing , 2020, ArXiv.

[10]  Qiang Yang,et al.  SecureBoost: A Lossless Federated Learning Framework , 2019, IEEE Intelligent Systems.

[11]  Qi Zhang,et al.  A Fast Algorithm for Approximate Quantiles in High Speed Data Streams , 2007, 19th International Conference on Scientific and Statistical Database Management (SSDBM 2007).

[12]  Yang Liu,et al.  Boosting Privately: Privacy-Preserving Federated Extreme Boosting for Mobile Crowdsensing , 2019, ArXiv.

[13]  Gaël Varoquaux,et al.  Scikit-learn: Machine Learning in Python , 2011, J. Mach. Learn. Res..

[14]  Haoyi Xiong,et al.  SecureGBM: Secure Multi-Party Gradient Boosting , 2019, 2019 IEEE International Conference on Big Data (Big Data).

[15]  Sanjeev Khanna,et al.  Space-efficient online computation of quantile summaries , 2001, SIGMOD '01.

[16]  Li Wang,et al.  A Hybrid-Domain Framework for Secure Gradient Tree Boosting , 2020, ArXiv.

[17]  Gerardo Ganis,et al.  Large-Scale Merging of Histograms using Distributed In-Memory Computing , 2015 .

[18]  Homin K. Lee,et al.  DDSketch: A Fast and Fully-Mergeable Quantile Sketch with Relative-Error Guarantees , 2019, Proc. VLDB Endow..

[19]  Jie Xu,et al.  The Tradeoff Between Privacy and Accuracy in Anomaly Detection Using Federated XGBoost , 2019, ArXiv.