Figure 1: Pruning unstable random matrices.
From: Symmetric and Asymmetric Tendencies in Stable Complex Systems
Results of the stabilization algorithm employed on 50,000 unstable 20 × 20 random matrices for the three different stabilization strategies (random, variance-minimizing and-minimizing and χoff-minimizing) described in the main text. Data points at each iteration indicate the sample average over the 50,000 simulations. Standard error of the mean estimates are on the order of 10−3 for both figures. (a) The maximum real eigenvalue at the end of each iteration. (b) The proportion of all matrices with a decreased maximum real eigenvalue from the previous iteration number.
