FORBID: Fast Overlap Removal by Stochastic GradIent Descent for Graph Drawing

While many graph drawing algorithms consider nodes as points, visualization tools often represent them shapes. These shapes support the display of information such labels or encode various data with size color. However, they can create overlaps between which hinder exploration process by hiding parts information. It is therefore utmost importance to remove these improve readability. If not handled layout process, Overlap Removal (OR) have been proposed post-processing. As layouts usually convey about their topology, it important that OR preserve much possible. We propose a novel algorithm models joint stress and scaling optimization problem, leverages efficient stochastic gradient descent. This approach compared state-of-the-art algorithms, several quality metrics demonstrate its efficiency quickly while retaining initial structures.