| Crates.io | nmi |
| lib.rs | nmi |
| version | 0.1.1 |
| created_at | 2025-06-14 15:42:54.992055+00 |
| updated_at | 2025-06-14 15:42:54.992055+00 |
| description | normalized mutual information |
| homepage | |
| repository | https://github.com/jean-pierreBoth/coreset |
| max_upload_size | |
| id | 1712442 |
| size | 44,048 |
Normalized Mutual information is based on the paper:
- Vinh.N.X Information Theoretic Measures for clustering comparison. Vinh 2010
This crate defines various quality measures of clustering based on information theory.
They are normalized with value in the interval [0,1] and some are metrics.
It is also possible to compare different clustering and to quantify how the affectation to items of 2 algorithms are related. It relies on a contingency table.
The various measures are described in the contingency file.
We preferentially use get_nmi_sqrt()
The trait affectation is used to make the contingency table aware of of data versus label association.
Licensed under either of
at your option.