cvi.modules.CONN

Connectivity-based CONN Cluster Validity Index.

This implementation follows the CONN-style validity index for prototype-based partitions. Unlike distance-only CVIs, CONN depends on the first and second best matching prototypes associated with each sample.

Notes

Incremental mode uses FuzzyART and assumes samples are already normalized to the ART input domain, typically [0, 1]. Batch mode supports FuzzyART, KMeans, and MiniBatchKMeans and can optionally normalize the full dataset before processing.

The iCONN initialization rule is handled explicitly:

  1. The first sample creates the first ART category.

  2. The second sample forces creation of the second ART category by temporarily setting ART vigilance to 1.0.

  3. Subsequent samples use ordinary ART dynamics.

References

    1. Merényi, “A new cluster validity index for prototype based clustering algorithms based on inter-and intra-cluster density,” 2007 International Joint Conference on Neural Networks, 2007.

    1. Tasdemir and E. Merényi, “A validity index for prototype-based clustering of data sets with complex cluster structures,” IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 41, no. 4, pp. 1039-1053, 2011.

      1. Brito da Silva, N. M. Melton, and D. C. Wunsch II, “Incremental cluster validity indices for online learning of hard partitions: Extensions and comparative study,” IEEE Access, vol. 8, pp. 22025-22047, 2020.

Module Attributes