FedTimeGAN: Generating Synthetic Time Series Data via Federated Learning of Generative Adversarial Networks
Abstract
Monitoring using accelerometers helps evaluate the quality of human activity. As the data collected is typically time series, dependent on the size and consistency of the data, it may be necessary to use synthetic data generation techniques. Data privacy is essential for personal user data, even synthetic data, and sharing it to train models may not be appropriate. Federated learning allows you to train models without sharing data. This work proposes a model for generating synthetic time series data for human activity via federated learning and Generative Adversarial Networks (GANs), denominated FedTimeGAN. A comparative analysis was conducted between federated and centralized training for the model of the synthetic time series data generation. The results reveal that federated learning can be more effective in generating high-quality synthetic data for this application, with data from multiple users and maintaining privacy.
References
Li, W., Chen, J., Wang, Z., Shen, Z., Ma, C., and Cui, X. (2023). Ifl-gan: Improved federated learning generative adversarial network with maximum mean discrepancy model aggregation. IEEE Transactions on Neural Networks and Learning Systems, 34:10502–10515.
McMahan, B., Moore, E., Ramage, D., Hampson, S., and y Arcas, B. A. (2017). Communication-efficient learning of deep networks from decentralized data. In Artificial intelligence and statistics, pages 1273–1282. PMLR.
Mo, L., Liu, S., Gao, R. X., John, D., Staudenmayer, J. W., and Freedson, P. S. (2012). Wireless design of a multisensor system for physical activity monitoring. IEEE transactions on Biomedical Engineering, 59(11):3230–3237.
Ranja, F., Nababan, E. B., and Candra, A. (2023). Synthetic data generation using time-generative adversarial network (time-gan) to predict cash atm. In 2023 International Conference on Computer, Control, Informatics and its Applications (IC3INA), pages 418–423. IEEE.
Reiss, A. (2012). PAMAP2 Physical Activity Monitoring. UCI Machine Learning Repository. DOI: 10.24432/C5NW2H.
Tolstikhin, I. O., Sriperumbudur, B. K., and Schölkopf, B. (2016). Minimax estimation of maximum mean discrepancy with radial kernels. Advances in Neural Information Processing Systems, 29.
Xin, B., Yang, W., Geng, Y., Chen, S., Wang, S., and Huang, L. (2020). Private flgan: Differential privacy synthetic data generation based on federated learning. pages 2927–2931.
Yoon, J., Jarrett, D., and Van der Schaar, M. (2019). Time-series generative adversarial networks. Advances in neural information processing systems, 32.
Zhang, Y., Zhang, Q., Yang, M., Xiao, T., and Wang, Z. (2023). Fedgan: Federated gan for few-shot image generation. pages 1020–1024. Institute of Electrical and Electronics Engineers Inc.
