All quantization bins are equally-sized. However, unless the values are uniformly distributed, uniform quantization is not optimal. Unfortunately, neural network weights are not uniform but rather have bell-shaped distributions, which is why non-uniform quantization is often used
Uniform Quantization
Creator
Creator
Seonglae ChoCreated
Created
2023 Oct 29 4:29Editor
Editor
Seonglae ChoEdited
Edited
2023 Oct 29 4:58Refs
Refs