Collaborative Random Forests Learning
Yohan Foucade, Younès Bennani, Zakaria Aabbou · 2021
Collaborative learning is an emerging field of machine learning. In this framework, multiple learning algorithms try to learn from a distributed database. The main idea is to improve the performance of each algorithm by exchanging information about the local structures of the data, without disclosing sensitive information. It is related to ensemble learning and federated learning. In this work, we presents a collaborative supervised learning approach. We used classification trees as a base algorithm, although the framework is generic and could be implemented with a wide class of algorithms. The process is divided into two steps: a local and a collaboration step. The local step uses a classical supervised learning algorithm, locally and independently on each dataset, which will result in obtaining multiple classifiers for each of those bases. During the collaboration step, each algorithm will collaborate with its remote counterparts. Our approach has been tested on several datasets and has shown promising results.