Processes, Год журнала: 2025, Номер 13(5), С. 1588 - 1588

Опубликована: Май 20, 2025

We propose a hierarchical federated learning (HFL) framework for predictive pollutant analysis in advanced green analytical chemistry (AGAC), addressing the limitations of centralized approaches scalability and data privacy. The system integrates localized sub-models with hybrid neural architectures, combining LSTM attention mechanisms to capture temporal dependencies feature importance distributed data, while raw measurements remain decentralized. A global aggregator dynamically adjusts model weights based on validation performance heterogeneity, ensuring robust adaptation diverse environmental conditions. interfaces seamlessly AGAC infrastructure, processing inputs from instruments into standardized sequences mapping predictions back concentrations through calibration curves. Implemented PyTorch Federated edge-cloud deployment, employs homomorphic encryption secure transmission, prioritizing spectral features critical organic detection. Our approach achieves superior accuracy privacy preservation compared traditional methods, offering transformative solution scalable monitoring. proposed method demonstrates significant potential real-world applications, particularly scenarios requiring collaboration without compromising integrity.

Язык: Английский