A Federated Data-driven Multiobjective Evolutionary Algorithm via Continual Learnable Clustering

Takato Kinoshita, Naoki Masuyama, Yusuke Nojima

Abstract: As solution evaluation costs increase, data-driven multiobjective optimization is becoming more critical. Using multiple computers in parallel helps manage computation times. Federated learning offers a privacy-friendly, cost-effective way to handle distributed data. Integration of these two approaches is known as federated data-driven multiobjective optimization. However, to the best of our knowledge, a few studies tackle this emerging topic. This paper introduces multiobjective evolutionary algorithms (MOEAs) to federated clustering via adaptive resonance theory-based clustering (FCAC) and proposes FCAC-MOEA as a solver system of federated data-driven optimization problems. The computational experiments showed that the proposed method achieves both high search efficiency and privacy preservation on various multiobjective optimization problems.