/this-paper-could-be-a-tweet

List of (mostly ML) papers, where description of the method could be shortened significantly

This paper could be a tweet

This is a list of (mostly ML) papers, where description of the method contains a lot of fluff, equation theatre and it could be shortened significantly and explained much better.

This does not mean that idea in paper is bad or results of mentioned papers are worthless. It just means, that in my opinion, they could be presented in much better fashion.

GhostNet: More Features from Cheap Operations

Idea is to replace (Pytorch pseudocode follow):

Conv2d(in, out, kernel_size)

With:

Sequential(
  Conv2d(in, small, kernel_size),
  Conv2d(small, out, kernel_size2, groups=small)
)

Aka factorized convolution in yet another way using smaller convolution + depthwise convolution.

Monarch: Expressive Structured Matrices for Efficient and Accurate Training

Instead of a bad figure and important piece of algorithm hidden in the middle of the page:

image

We could have much better figure (parts taken from Shufflenet):

image

With this, paper could be understood in seconds instead of hours.