Variational Autoencoders for Highly Multivariate Spatial Point Processes Intensities
Abstract
Multivariate spatial point process models can describe heterotopic data over space. However, highly multivariate intensities are computationally challenging due to the curse of dimensionality. To bridge this gap, we introduce a declustering based hidden variable model that leads to an efficient inference procedure via a variational autoencoder (VAE). We also prove that this model is a generalization of the VAE-based model for collaborative filtering. This leads to an interesting application of spatial point process models to recommender systems. Experimental results show the method's utility on both synthetic data and real-world data sets.
Cite
Text
Yuan et al. "Variational Autoencoders for Highly Multivariate Spatial Point Processes Intensities." International Conference on Learning Representations, 2020.Markdown
[Yuan et al. "Variational Autoencoders for Highly Multivariate Spatial Point Processes Intensities." International Conference on Learning Representations, 2020.](https://mlanthology.org/iclr/2020/yuan2020iclr-variational/)BibTeX
@inproceedings{yuan2020iclr-variational,
title = {{Variational Autoencoders for Highly Multivariate Spatial Point Processes Intensities}},
author = {Yuan, Baichuan and Wang, Xiaowei and Ma, Jianxin and Zhou, Chang and Bertozzi, Andrea L. and Yang, Hongxia},
booktitle = {International Conference on Learning Representations},
year = {2020},
url = {https://mlanthology.org/iclr/2020/yuan2020iclr-variational/}
}