Serverless gossip training of LSTM failure detectors: A matched-protocol comparison with federated, local and centralized learning on NASA C-MAPSS
Abstract
Industrial predictive maintenance increasingly depends on learning from equipment spread across sites whose sensor data cannot easily be pooled. Federated averaging (FedAvg) solves this with a central aggregation server; gossip learning removes the server, but its behaviour for recurrent failure-detection models has not been measured under controlled conditions. We compare synchronous ring gossip with FedAvg, isolated local training and a centralized reference for a stacked LSTM that detects imminent failure on the NASA C-MAPSS turbofan benchmark. All methods share one open implementation, architecture, initialization, optimizer, data split and training budget, and the primary endpoint uses one terminal window per test engine to avoid the statistical dependence of overlapping windows. On FD001 (five seeds), gossip reached a terminal-window F1 of 89.6 +/- 1.3%, compared with 89.9 +/- 1.1% for FedAvg, 83.6 +/- 6.7% for local training and 93.5 +/- 2.1% for centralized training, while transmitting the same payload as FedAvg without a coordinator. Node models agreed closely but not exactly (1.8% pairwise decision disagreement versus 5.6% without communication). Across FD002-FD004, peer communication improved terminal-window F1 over local training by 13-28 points; gossip matched FedAvg on FD003 and FD004 but was 4.3 points lower on the multi-condition FD002 subset. Simulated message loss, node failure and server outage changed neither method appreciably, whereas larger rings degraded gossip faster. Ring gossip is therefore a practical serverless alternative when data heterogeneity is moderate, and faster-mixing topologies become important as heterogeneity grows.
Related
- Proof of Stake economy under centralized exchanges--a mean field model
- Relaxed Centralized Circumcentered-Reflection Methods for Convex Feasibility: Convergence and Finite Termination
- Centralized Fault-Location and PMU Placement Strategies in Weakly Observable Grids: A Comprehensive Review
- Exact Regret Frontiers and Externality Scheduling in Centralized Serial-Dictatorship Bandits
- Centralized Multi-Objective UAV–UGV Task Allocation for Disaster Relief Material Distribution with Mobile UAV Bases
BibTeX
@misc{ozturk2026serverless,
title = {{Serverless gossip training of LSTM failure detectors: A matched-protocol comparison with federated, local and centralized learning on NASA C-MAPSS}},
author = {Yusuf Öztürk and Enes Göktekin and Bengisu Atlı and Akın Öztürk and Zhixiang Wang and Ulas Bagci},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.35792},
eprint = {2609.35792},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.35792},
note = {CC BY-NC-ND 4.0},
}
RIS
TY - UNPB
TI - Serverless gossip training of LSTM failure detectors: A matched-protocol comparison with federated, local and centralized learning on NASA C-MAPSS
AU - Yusuf Öztürk
AU - Enes Göktekin
AU - Bengisu Atlı
AU - Akın Öztürk
AU - Zhixiang Wang
AU - Ulas Bagci
PY - 2026
DA - 2026/09/17
PB - arXiv
DO - 10.48550/arxiv.2609.35792
UR - https://arxiv.org/abs/2609.35792
AB - Industrial predictive maintenance increasingly depends on learning from equipment spread across sites whose sensor data cannot easily be pooled. Federated averaging (FedAvg) solves this with a central aggregation server; gossip learning removes the server, but its behaviour for recurrent failure-detection models has not been measured under controlled conditions. We compare synchronous ring gossip with FedAvg, isolated local training and a centralized reference for a stacked LSTM that detects imminent failure on the NASA C-MAPSS turbofan benchmark. All methods share one open implementation, architecture, initialization, optimizer, data split and training budget, and the primary endpoint uses one terminal window per test engine to avoid the statistical dependence of overlapping windows. On FD001 (five seeds), gossip reached a terminal-window F1 of 89.6 +/- 1.3%, compared with 89.9 +/- 1.1% for FedAvg, 83.6 +/- 6.7% for local training and 93.5 +/- 2.1% for centralized training, while transmitting the same payload as FedAvg without a coordinator. Node models agreed closely but not exactly (1.8% pairwise decision disagreement versus 5.6% without communication). Across FD002-FD004, peer communication improved terminal-window F1 over local training by 13-28 points; gossip matched FedAvg on FD003 and FD004 but was 4.3 points lower on the multi-condition FD002 subset. Simulated message loss, node failure and server outage changed neither method appreciably, whereas larger rings degraded gossip faster. Ring gossip is therefore a practical serverless alternative when data heterogeneity is moderate, and faster-mixing topologies become important as heterogeneity grows.
ER -