Subspace least-squares multidimensional scaling
https://arxiv.org/abs/1709.03484
© Amit Boyarski, Adi Weinberger, 2020
The code solves a stress based MDS problem where the input is a triangular mesh. It uses the SMACOF algorithm with a subspace parametrization based on the eigedecomposition of the Laplace-Beltrami operator. This parametrization allows solving huge MDS problems in a fraction of the time compared to the standard SMACOF algorithm.
In order to run the project, you should put a 'shape_name.off' file, and a pre-computed matrix of pairwise (geodeisc) distances 'D_shape_name.mat' in the 'input' folder, and update the paths in the main.py file. Then run 'main.py' and it will generate the embedding.