gradient compression and quantization
Used in 2 articles
Gradient compression and quantization are techniques used to reduce the amount of data needed to transfer information during training of machine learning models, which can help protect patient privacy by minimizing the amount of sensitive data that needs to be transmitted.