Ensemble Weighting Strategy for Federated Learning to Handle Heterogeneous Data Distributions.
- Resource Type
- Article
- Authors
- Richter, Lucas; Dontsov, Ilja; Jacob, Tania
- Source
- IUP Journal of Electrical & Electronics Engineering. Oct2022, Vol. 15 Issue 4, p7-20. 14p.
- Subject
- *DATA distribution
*LEARNING strategies
*SMART cities
*CLASSIFICATION algorithms
*DATA security
- Language
- ISSN
- 2583-519X
Measured data in the context of smart cities can be used to develop new and innovative business models to improve efficiency and the value of life. A time-series classification algorithm can help automatize many different processes such as forecasting services. In order to ensure data security and privacy, Federated Learning trains a global model collaboratively on multiple clients. Having different data-distributions and data-quantities across participating clients, neural networks suffer from slow convergence and overfitting. Based on different data values and number of clients, different dataclustering strategies have been developed and evaluated in this study to update global model weights. Public time-series data has been downloaded from the Internet to generate various synthetic datasets and train a Relational-Regularized Autoencoder for classification purposes. The results showed an improvement of model performance concerning generalization. [ABSTRACT FROM AUTHOR]