![[2602.21928] Learning Unknown Interdependencies for Decentralized Root Cause Analysis in Nonlinear Dynamical Systems](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.21928] Learning Unknown Interdependencies for Decentralized Root Cause Analysis in Nonlinear Dynamical Systems
Summary
This paper presents a novel federated learning methodology for decentralized root cause analysis in nonlinear dynamical systems, addressing the challenges of unknown interdependencies among distributed clients.
Why It Matters
Root cause analysis (RCA) is critical in complex industrial systems, where understanding interdependencies can significantly improve operational efficiency and cybersecurity. This research provides a framework that enhances RCA without compromising client data privacy, making it highly relevant for industries reliant on IoT and machine learning.
Key Takeaways
- Introduces a federated learning approach for decentralized RCA.
- Addresses the challenge of heterogeneous data features among clients.
- Ensures privacy through calibrated differential privacy noise.
- Validates methodology with simulations and real-world datasets.
- Establishes theoretical convergence guarantees for the proposed model.
Computer Science > Machine Learning arXiv:2602.21928 (cs) [Submitted on 25 Feb 2026] Title:Learning Unknown Interdependencies for Decentralized Root Cause Analysis in Nonlinear Dynamical Systems Authors:Ayush Mohanty, Paritosh Ramanan, Nagi Gebraeel View a PDF of the paper titled Learning Unknown Interdependencies for Decentralized Root Cause Analysis in Nonlinear Dynamical Systems, by Ayush Mohanty and 2 other authors View PDF HTML (experimental) Abstract:Root cause analysis (RCA) in networked industrial systems, such as supply chains and power networks, is notoriously difficult due to unknown and dynamically evolving interdependencies among geographically distributed clients. These clients represent heterogeneous physical processes and industrial assets equipped with sensors that generate large volumes of nonlinear, high-dimensional, and heterogeneous IoT data. Classical RCA methods require partial or full knowledge of the system's dependency graph, which is rarely available in these complex networks. While federated learning (FL) offers a natural framework for decentralized settings, most existing FL methods assume homogeneous feature spaces and retrainable client models. These assumptions are not compatible with our problem setting. Different clients have different data features and often run fixed, proprietary models that cannot be modified. This paper presents a federated cross-client interdependency learning methodology for feature-partitioned, nonlinear time-series...
