Training Large-Vocabulary Neural Language Models by Private Federated Learning for Resource-Constrained Devices
This addresses privacy-preserving language model training for resource-constrained devices, but it is incremental as it builds on existing FL and DP methods with optimizations.
The paper tackles the problem of training large neural network language models on resource-constrained devices using private federated learning, where differential privacy noise hinders convergence; the result is a combination of techniques like Partial Embedding Updates, LoRA, and NCE that enable training while preserving accuracy and privacy.
Federated Learning (FL) is a technique to train models using data distributed across devices. Differential Privacy (DP) provides a formal privacy guarantee for sensitive data. Our goal is to train a large neural network language model (NNLM) on compute-constrained devices while preserving privacy using FL and DP. However, the DP-noise introduced to the model increases as the model size grows, which often prevents convergence. We propose Partial Embedding Updates (PEU), a novel technique to decrease noise by decreasing payload size. Furthermore, we adopt Low Rank Adaptation (LoRA) and Noise Contrastive Estimation (NCE) to reduce the memory demands of large models on compute-constrained devices. This combination of techniques makes it possible to train large-vocabulary language models while preserving accuracy and privacy.