Normalized mutual information is a biased measure for classification and community detection.

Maximilian Jerdee1,2, Alec Kirkley3,4,5, Mark Newman6,7

  • 1Department of Physics, University of Michigan, Ann Arbor, MI, USA.

Nature Communications
|December 11, 2025
PubMed
Summary

Normalized mutual information (NMI) is biased in clustering and classification evaluations. This study introduces a modified NMI to correct these biases, significantly impacting conclusions on algorithm performance, especially for network community detection.

Related Concept Videos

Mutual Inductance01:24

Mutual Inductance

Inductance is the property of a device that tells us how effectively it induces an emf in another device. In other words, it is a physical quantity that expresses the effectiveness of a given device.
When two circuits carrying time-varying currents are close to one another, the magnetic flux through each circuit varies because of the changing current in the other circuit. Consequently, an emf is induced in each circuit by the changing current in the other. Therefore, this type of emf is called...
3.5K
Confidence Coefficient01:24

Confidence Coefficient

The confidence coefficient is also known as the confidence level or degree of confidence. It is the percent expression for the probability, 1-α, that the confidence interval contains the true population parameter assuming that the confidence interval is obtained after sufficient unbiased sampling; for example, if the CL = 90%, then in 90 out of 100 samples the interval estimate will enclose the true population parameter. Here α is the area under the curve, distributed equally under...
10.3K
Midrange01:07

Midrange

A somewhat easy to compute quantitative estimate of a data set’s central tendency is its midrange, which is defined as the mean of the minimum and maximum values of an ordered data set.
Simply put, the midrange is half of the data set’s range. Similar to the mean, the midrange is sensitive to the extreme values and hence the prospective outliers. However, unlike the mean, the midrange is not sensitive to all the values of the data set that lie in the middle. Thus, it is prone to...
4.1K
Test for Homogeneity01:23

Test for Homogeneity

The goodness–of–fit test can be used to decide whether a population fits a given distribution, but it will not suffice to decide whether two populations follow the same unknown distribution. A different test, called the test for homogeneity, can be used to conclude whether two populations have the same distribution. To calculate the test statistic for a test for homogeneity, follow the same procedure as with the test of independence. The hypotheses for the test for homogeneity can...
2.3K
Kendall's Coefficient of Concordance01:20

Kendall's Coefficient of Concordance

Kendall's Coefficient of Concordance (W), also known as Kendall's W, is a non-parametric statistical measure used to assess the agreement or concordance between multiple raters or judges when they rank a set of items. It is often used when you have ordinal data (ranks) and you want to see if there is consistency or consensus among the raters. It is widely applied in research areas such as psychology, medicine, and social sciences, where multiple judges are asked to rank or rate subjects...
915
Aggregates Classification01:29

Aggregates Classification

Aggregate classification is generally based on its size, petrographic characteristics, weight, and source. Size classification ranges from coarse to fine aggregates, defined by the size of the particles. Coarse aggregates are particles that do not pass through ASTM sieve No. 4, and aggregates that pass through the sieve are fine aggregates.
Petrographic classification groups aggregates based on common mineralogical characteristics. Some of the common mineral groups found in aggregates are...
950