nov 20, 2018 - Bichen Wu: Mixed precision quantization of convnets via differentiable neural architecture search, 2018
Description:
首次提出了逐层混合精度的想法，讲不通的层量化为不同的精度。本文将该问题看作神经结构搜索问题，并提出DNAS框架，根据基于梯度的优化，进行高效的解空间搜索。Added to timeline:
Date:
nov 20, 2018
Now
~ 6 years and 7 months ago