Computation of Pixel-Oriented Grid Layout for 2D Datasets Using VRGrid

The arrangement of n points in a \(\sqrt{n}\times \sqrt{n}\) sized grid offers an efficient and overlap-free way to visualize data. By making use of the Voronoi Relaxation method, we propose a novel post-processing algorithm called VRGrid which allows the arrangement of any two-dimensional data in a grid while minimizing disformation of the input data. This method can be used with popular but overlap-prone projection methods such as t-SNE or MDS to obtain overlap-free and compact visualizations of data. In this chapter, we present how VRGrid works and its complexity, and benchmark it against the state-of-the-art methods Self-Sorting Maps and Distance-preserving Grid using several metrics to measure the quality of the obtained outputs. Given sufficient processing time, VRGrid outperforms these methods in preserving points pairwise-distances and minimizing the disformation of the input data.