FedGA-Tree: Federated Decision Tree using Genetic Algorithm

--In recent years, with rising concerns for data privacy, Federated Learning has gained prominence, as it enables collaborative training without the aggregation of raw data from participating clients. However, much of the current focus has been on parametric gradient-based models, while nonparametric counterparts such as decision tree are relatively understudied. Existing methods for adapting decision trees to Federated Learning generally combine a greedy tree-building algorithm with differential privacy to produce a global model for all clients. These methods are limited to classification trees and categorical data due to the constraints of differential privacy. In this paper, we explore an alternative approach that utilizes Genetic Algorithm to facilitate the construction of personalized decision trees and accommodate categorical and numerical data, thus allowing for both classification and regression trees. Comprehensive experiments demonstrate that our method surpasses decision trees trained solely on local data and a benchmark algorithm. With rapid advancement of AI and machine learning, there are many concerns about data usage and privacy. Lawmakers worldwide have attempted to create incentives for companies to focus more on privacy in their model development, with key examples including the General Data Protection Regulations implemented by the European Union and the California Consumer Privacy Act.Federated Learning (FL) was introduced by Google as an approach for mobile devices to collaboratively solve a machine learning problem without sharing user's local data [14], [17]. In the FL framework, multiple clients contribute to solve a machine learning problem while maintaining their data locally. A global server helps aggregate information that clients deem fit to share, such as model weights, and construct an improved model. The two main scenarios of data distribution in FL are horizontal and vertical. In the former, clients have the same features but different set of samples while in the latter, clients have different features but the same set of samples. Currently, the main focus of the FL research community is on parametric, gradient-based models, yet there is an expanding body of literature that explores the use of decision tree models [25], [26] [7], [19].