Federated Learning of Dynamic Bayesian Network via Continuous Optimization from Time Series Data

2024-12-13

AI Summary

Privacy preservation and statistical heterogeneity pose significant challenges for dynamic Bayesian network (DBN) structure learning from distributed time-series data in federated settings. Method: We propose the first continuous-optimization-based federated DBN structure learning framework. To address inter-client statistical heterogeneity, we design PFDBNL—a personalized regularization method that integrates proximal operator constraints with parameter-level model exchange, enabling collaborative modeling across both homogeneous and heterogeneous horizontally partitioned clients while ensuring data locality and privacy. Contribution/Results: PFDBNL achieves efficient and robust estimation of the global DBN structure without centralized data access. Extensive experiments on synthetic and real-world datasets demonstrate that our method significantly outperforms existing baselines, particularly in typical federated scenarios characterized by a large number of clients and limited local samples—yielding substantial improvements in structural recovery accuracy.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Constraint Satisfaction and Optimization: Distributed CSP/Optimization
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Systems and Infrastructure for Web, Mobile and WoT: Federated Web and WoT systems, including distributed, federated and edge-based data processing

Abstract

Traditionally, learning the structure of a Dynamic Bayesian Network has been centralized, requiring all data to be pooled in one location. However, in real-world scenarios, data are often distributed across multiple entities (e.g., companies, devices) that seek to collaboratively learn a Dynamic Bayesian Network while preserving data privacy and security. More importantly, due to the presence of diverse clients, the data may follow different distributions, resulting in data heterogeneity. This heterogeneity poses additional challenges for centralized approaches. In this study, we first introduce a federated learning approach for estimating the structure of a Dynamic Bayesian Network from homogeneous time series data that are horizontally distributed across different parties. We then extend this approach to heterogeneous time series data by incorporating a proximal operator as a regularization term in a personalized federated learning framework. To this end, we propose exttt{FDBNL} and exttt{PFDBNL}, which leverage continuous optimization, ensuring that only model parameters are exchanged during the optimization process. Experimental results on synthetic and real-world datasets demonstrate that our method outperforms state-of-the-art techniques, particularly in scenarios with many clients and limited individual sample sizes.

Problem

• Federated learning for Dynamic Bayesian Networks
• Handling data heterogeneity in distributed systems
• Privacy-preserving collaborative structure learning

Innovation

• Federated learning for Dynamic Bayesian Network
• Proximal operator for data heterogeneity
• Continuous optimization in personalized framework