We demonstrate that a production-quality keyword-spotting model can be trained on-device using federated learning and achieve comparable false accept and false reject rates to a centrally-trained model. To overcome the algorithmic constraints associated with fitting on-device data (which are inherently non-independent and identically distributed), we conduct thorough empirical studies of optimization algorithms and hyperparameter configurations using large-scale federated simulations. To overcome resource constraints, we replace memory intensive MTR data augmentation with SpecAugment, which reduces the false reject rate by 56%. Finally, to label examples (given the zero visibility into on-device data), we explore teacher-student training.
@article{arxiv.2005.10406,
title = {Training Keyword Spotting Models on Non-IID Data with Federated Learning},
author = {Andrew Hard and Kurt Partridge and Cameron Nguyen and Niranjan Subrahmanya and Aishanee Shah and Pai Zhu and Ignacio Lopez Moreno and Rajiv Mathews},
journal= {arXiv preprint arXiv:2005.10406},
year = {2020}
}