WEHD - Weighted Euclidean-Hamming Distance

A heterogeneous distance function for use in scientific Python environments. The weights for an optimal metric for a dataset can be discovered using gradient-free optimizers, such as Evolution Strategies, in unsupervised settings, as demonstrated in this repo.

References

Wilson, D. R., & Martinez, T. R. (1997). Improved heterogeneous distance functions. Journal of artificial intelligence research, 6, 1-34.

Gupta, A. A., Foster, D. P., & Ungar, L. H. (2008). Unsupervised distance metric learning using predictability. Technical Reports (CIS), 885.

Li, C., & Li, H. (2010). A Survey of Distance Metrics for Nominal Attributes. J. Softw., 5(11), 1262-1269.

Shi, Y., Li, W., & Sha, F. (2016, March). Metric learning for ordinal data. In Proceedings of the AAAI Conference on Artificial Intelligence (Vol. 30, No. 1).