So you've got 118 elements with ideal locations and a whole bunch of rhombi with known locations. The Euclidean distance (without bothering with the sqrt) is used as the "cost" function for the Hungarian algorithm (https://en.wikipedia.org/wiki/Hungarian_algorithm): the idea is to minimize the cost (distance from ideal) for each element.
You will sometimes see some gaps in the periodic table as this layout isn't "perfect" (in the sense of without isolated rhombi inside the table) and I did work on having a post-layout pass where these islands were identified and filled by moving adjacent rhombus assignments, but I wasn't happy with how this tended to break the layout of the periodic table (I was going for some visual "looks similar to the classic U shaped periodic table) and so decided to accept them.