mlsanityData to 7 Oct 2026

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

arXiv 2307.01282 ↗First posted 3 Jul 2023cs.SI stat.MLcited by 7

Not on the map. Why?

Abstract (arXiv)

Normalized mutual information is widely used as a similarity measure for evaluating the performance of clustering and classification algorithms. In this paper, we argue that results returned by the normalized mutual information are biased for two reasons: first, because they ignore the information content of the contingency table and, second, because their symmetric normalization introduces spurious dependence on algorithm output. We introduce a modified version of the mutual information that remedies both of these shortcomings. As a practical demonstration of the importance of using an unbiased measure, we perform extensive numerical tests on a basket of popular algorithms for network community detection and show that one's conclusions about which algorithm is best are significantly affected by the biases in the traditional mutual information.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Builds on 15 of its 69 references that are in the map. First posted on arXiv 3 Jul 2023.

Builds on

15 of 69 references.

References · 15

  1. 2018From Louvain to Leiden: guaranteeing well-connected communitiescited by 379
  2. 2016Community detection in networks: A user guidecited by 105
  3. 2013Community detection thresholds and the weak Ramanujan propertycited by 47
  4. 2016The ground truth about metadata and community detection in networkscited by 42
  5. 2017The Impact of Random Models on Clustering Similaritycited by 34
  6. 2016A Comparative Analysis of Community Detection Algorithms on Artificial Networkscited by 23
  7. 2016Nonparametric Bayesian inference of the microcanonical stochastic block modelcited by 21
  8. 2018Prioritizing network communitiescited by 14
  9. 2022Inference of hyperedges and overlapping communities in hypergraphscited by 13
  10. 2015Evaluating accuracy of community detection using the relative normalized mutual informationcited by 7
  11. 2020Revealing consensus and dissensus between network partitionscited by 7
  12. 2023Network community detection via neural embeddingscited by 7
3 more
  1. 2019Improved mutual information measure for classification and community detectioncited by 6
  2. 2022Bayan Algorithm: Detecting Communities in Networks Through Exact and Approximate Optimization of Modularitycited by 5
  3. 2013A primer on information theory, with applications to neurosciencecited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 1; 2025: 3; 2026: 2.

7 papers.

  1. 2024Dial-In LLM: Human-Aligned LLM-in-the-loop Intent Clustering for Customer Service Dialoguescited by 11
  2. 2023Analyzing Modularity Maximization in Approximation, Heuristic, and Graph Neural Network Algorithms for Community Detectioncited by 4
  3. 2026Matrix Factorization Framework for Community Detection under the Degree-Corrected Block Modelcited by 1
  4. 2025Robust Markov stability for community detection at a scale learned based on the structurecited by 0
  5. 2025Assessing wildfire susceptibility in Iran: Leveraging machine learning for geospatial analysis of climatic and anthropogenic factorscited by 0
  6. 2025Unifying Information-Theoretic and Pair-Counting Clustering Similaritycited by 0
  7. 2026NMINE: Normalized Mutual Information Neural Estimationcited by 0