Federated Learning on Wearable Devices

This is the implementation of the paper: Demo Abstract: Federated Learning on Wearable Devices.

Requirements

• Java 11

Data

We setup our demo in a controlled environment with a Human Activity Recognition (HAR) dataset. The dataset is built from a Daily and Sports Activities Data Set (DSA), which comprises of motion sensor data of 19 daily sports activities each performed by 8 subjects in their own style for 5 minutes.

Model

The selected model is a MLP, composed of one input and one output layer, and one hidden layer with 1000 units using ReLU activations.

Run

Results

Ackonwledgements

This work is sponsored by CERNET Innovation Project (NGII20190105) and Academy of Finland grant 325774 and 325570.

References

[1] Billur Barshan and Murat Cihan Yüksek. 2013. Recognizing Daily and Sports Activities in Two Open Source Machine Learning Environments Using Body-Worn Sensor Units. The Computer Journal 57, 11 (2013), 1649–1667.

[2] Y. Chen, X. Qin, J. Wang, C. Yu, and W. Gao. 2020. FedHealth: A Federated Transfer Learning Framework for Wearable Healthcare. IEEE Intelligent Systems 35, 4 (2020), 83–93.

[3] H. Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. 2017. Communication-Efficient Learning of Deep Networks from Decentralized Data. In Proceedings of the 20th International Conference on Artificial Intelligence and Statistics.