Copying is a common problem in papers. Many papers are similar and need to be revised.

In my opinion, to reproduce the methods in the paper, we must meet the following conditions: fully understand the methods in the paper, including the overall structure, the composition of each module, how to set the parameters and superparameters of each part, and so on. If some details in the paper are not clear, you can send an email to the author to ask. Have their own familiar programming languages and corresponding packages or frameworks (such as Numpy, Scipy, Pytorch, etc.). ).

It is generally not easy to realize everything from scratch, but with the help of existing libraries and building blocks, it can be copied quickly.