Andrew Hard(Google Inc.), Kurt Partridge(Google Inc.), Cameron Nguyen(Google Inc.), Niranjan Subrahmanya(Google Inc.), Aishanee Shah(Google Inc.), Pai Zhu(Google Inc.), Ignacio Moreno(Google Inc.) and Rajiv Mathews(Google Inc.)
Abstract:
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.