This is a huge struggle for me as a computational biologist with a computer science background. When I use a published method I try to have at least a cursory understanding of how the method works. However, many of the papers in my field are now statistics and machine learning heavy. The lack of a standardized notation makes the method descriptions very hard for me to follow. Occasionally the source code published alone with the papers is readable and the implementation is much simpler than I expect. But it's not uncommon for the published code to be a pile of spaghetti with no comments or sensible variable names.