pVS-Functional-Diversity

This repository contains R code for pVS, a new measure of functional diversity based on the Vendi Score.

We provide code to reproduce all experiments shown in our paper (Reference Incoming). For additional details about the Vendi Score, the role of order q, and python implentation, check out the Vendi Score repository.

Comparisons against other Metrics

In our paper, we highlight how the pVS can handle a diverse set of scenarios for which many other metrics fail.

We use implementations of the tested metrics from the following sources

• fundiversity: Used for FDiv, FRic, FDis, Rao's Q
• BAT: Used for richness (alpha), dispersion, evenness. We construct trees using the ape package.
• All other metrics are computed using custom implentations in Experiments/DiversityMetrics.R. The TED and TOP metrics use the official implementations from their paper.

References

@article{friedman2023vendi,
  title={The Vendi Score: A Diversity Evaluation Metric for Machine Learning},
  author={Friedman, Dan and Dieng, Adji Bousso},
  journal={Transactions on Machine Learning Research},
  issn={2835-8856},
  year={2023}
}

@inproceedings{pasarkar2024cousins,
  title={Cousins Of The Vendi Score: A Family Of Similarity-Based Diversity Metrics For Science And Machine Learning},
  author={Pasarkar, Amey P and Dieng, Adji Bousso},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  pages={3808--3816},
  year={2024},
  organization={PMLR}
}