Fig. 2
From: Module-based regularization improves Gaussian graphical models when observing noisy data
Performance comparison of the Standard and the Modular GLASSO in detecting planted partitions under low and high noise conditions. The adjusted mutual information (AMI) between recovered and planted partitions shows that the Standard GLASSO finds the planted partition only if the samples are few when the noise level is high, but when samples and within-module covariance are sufficient for low noise. In contrast, the Modular GLASSO finds the planted partition also in high noise when samples and covariance are sufficient