Skip to main content
Fig. 2 | BMC Bioinformatics

Fig. 2

From: Gene regulation network inference using k-nearest neighbor-based mutual information estimation: revisiting an old DREAM

Fig. 2

Percent error of different mutual information estimators for multivariate gaussian distribution. Each boxplot represents 100 replicates, with columns representing sample size = {100,1K,10K}, and rows the correlation = {0.3,0.6,0.9}. A Percent error (y-axis) for two-way mutual information (MI2) was compared for 4 different methods: ML_Sq = Maximum Likelihood (Shannon’s MI) with fixed width binning (number of bins is determined by square-root), MM_Sq = Miller–Madow’s formula for MI with square-root for the number of bins, KL3 = KL formula for kNN-MI with k = 3, KSG3 = KSG formula for kNN-MI with k = 3; B same methods compared for total correlation (TC)

Back to article page

BMC Bioinformatics

ISSN: 1471-2105

Contact us