Information theoretic measures for clusterings comparison: is a correction for chance necessary?

By Xuan Vinh Nguyen, Julien Epps, James Bailey. International Conference on Machine Learning, 2009

1,194 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Signal Processing

View Paper: https://www.semanticscholar.org/paper/35c15d7dfa1fddd2c0292146412ebbec704e8be9

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Signal Processing

Citations in the corpus

Evaluating User Privacy in Bitcoin (2013, background): “NMI assesses the similarity of two groupings of the same items (in our case, Eprof and GTprof ), and takes higher values (1) the more identical the groupings are [19, 20].”
Analyzing Speaker Information in Self-Supervised Models to Improve Zero-Resource Speech Processing (2021, methodology): “By mapping each unit to the overlapping phone label in the forced alignment, we evaluate the clustering quality in terms of the adjusted rand index (ARI), adjusted mutual information (AMI) [26], homogeneity, and completeness [27].”
Which fMRI clustering gives good brain parcellations? (2014, methodology): “214 The adjusted Rand index (ARI) is comprised between -1 and 1, and measures the consistency of the two 215 labellings while being invariant to a permutation of the labels (Vinh et al., 2009).”
Conditional distribution variability measures for causality detection (2016, background): “It corrects the effect of agreement solely due to chance, Vinh et al. (2009).”
Stable Community Cores in Complex Networks (2012, methodology): “Also, it is shown that the mutual information depends on the size of the partitions [20], therefore we used an adjusted version of this metric, called AMI [20].”
The ground truth about metadata and community detection in networks (2016, background): “Other normalizing factors in the denominator are possible, including 12 [H(u)+H(v)] and max{H(u), H(v)}; see [58].”
Algorithms for hierarchical clustering: an overview, II (2017, background): “Also see Vinh et al.(82) for additional information regarding correction for chance agreement.”
Unsupervised Point Cloud Pre-training via Occlusion Completion (2020, methodology): “To evaluate the clustering quality, we calculate the adjusted mutual information (AMI) [33] between the generated and the ground-truth clusters.”
Human brain mapping: A systematic comparison of parcellation methods for the human cerebral cortex (2017, background): “Intuitively, N00 and N11 account for the agreement of parcellations, whereas N01 and N10 indicate their disagreement (Vinh et al., 2009).”
The effects of data quality on machine learning performance on tabular data (2022, neutral): “We use an adapted version of MI called the adjusted mutual information (AMI) [51].”
The Eﬀects of Data Quality on Machine Learning Performance (2022, methodology): “We use an adapted version of MI called the Adjusted Mutual Information (AMI) [51].”
Multiple Kernel Fuzzy Clustering (2012, neutral): “(11)”