My gut feeling is that yes, this is pretty much self evident.

However, the interesting part of the paper is that they use that equivalency to propose that properties of polynomial regression are applicable to Neural networks, and draw some conclusions from that.