Maximum Entropy Density Estimation with Incomplete Presence-Only Data
Abstract
We demonstrate a generalization of Maximum Entropy Density Estimation that elegantly handles incomplete presence-only data. We provide a formulation that is able to learn from known values of incomplete data without having to learn imputed values, which may be inaccurate. This saves the effort needed to perform accurate imputation while observing the principle of maximum entropy throughout the learning process. We provide analysis and examples of our algorithm under different settings of missing data.
Cite
Text
Huang and Salleb-Aouissi. "Maximum Entropy Density Estimation with Incomplete Presence-Only Data." Proceedings of the Twelfth International Conference on Artificial Intelligence and Statistics, 2009.Markdown
[Huang and Salleb-Aouissi. "Maximum Entropy Density Estimation with Incomplete Presence-Only Data." Proceedings of the Twelfth International Conference on Artificial Intelligence and Statistics, 2009.](https://mlanthology.org/aistats/2009/huang2009aistats-maximum/)BibTeX
@inproceedings{huang2009aistats-maximum,
title = {{Maximum Entropy Density Estimation with Incomplete Presence-Only Data}},
author = {Huang, Bert and Salleb-Aouissi, Ansaf},
booktitle = {Proceedings of the Twelfth International Conference on Artificial Intelligence and Statistics},
year = {2009},
pages = {240-247},
volume = {5},
url = {https://mlanthology.org/aistats/2009/huang2009aistats-maximum/}
}