Figure 1From: Automatic classification of protein structures using low-dimensional structure space mappingsConvergence of SMACOF with random and CMDS initializations. Left: initial stress is much greater with the random initialization and takes many more iterations of SMACOF to converge. Right: The Procrustes statistic between the random and CMDS initializations shows that their SMACOF coordinates become very similar as the number of iterations increase.Back to article page