Federated Learning in Indonesia: Challenges and Opportunities

Abstract

A single client of traditional Machine Learning (ML) limits the data sharing capabilities among different models. A centralized ML scheme leads to less data provision and confinement of a certain problem. In realistic applications, multiple learning clients need to work together to jointly train the model to enable a variety of solutions to scalability problems. However, the topic is relatively new. In this research, related open cases in a developing country are discussed, focusing on the exploration of emerging opportunities and challenges for the next generation of ML in the field of smart city, smart industry, and smart health service with case-studies from Indonesia. Federated Learning (FL) is a new paradigm of ML with a multi-client scheme. FL makes it possible to train ML algorithms with distributed training data across multiple devices or servers without any data exchange. Several large industries and companies in Indonesia were observed in this research to find out which sectors could most realize and increase the effectiveness and excellence of FL in advancing ML. Then possible solutions to deal with these challenges are proposed, one of which is the internet of things for smart cities and industries based on the federation mechanism which has great sector potential in Indonesia. Some areas of government and industry where FL can be integrated seamlessly are also listed. This research contributes in providing ideas for implementing FL in various fields to accelerate the development of FL in Indonesia. Finally, this research can briefly answer the question ‘What are the opportunities and challenges of FL in Indonesia?’.