This contains the core implementation of importance sampling algorithm in Mercury.
If you find this useful for your work, please consider citing:
@InProceedings{mercury2021sensys,
title = {Mercury: Efficient On-Device Distributed DNN Training via Stochastic Importance Sampling},
author = {Zeng, Xiao and Yan, Ming and Zhang, Mi},
booktitle = {ACM Conference on Embedded Networked Sensor Systems (SenSys)},
year = {2021}
}