.png){kind=icon}
.png){kind=icon}
Return
A VAE approach to sample multivariate extremes
.png){kind=icon}
PRE
AI
DOI：10.1080/00949655.2026.2645880.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Generating accurate extremes from an observational data set is crucial when seeking to estimate risks associated with the occurrence of future extremes which could be larger than those already observed. Applications range from the occurrence of natural disasters to financial crashes. Generative models from the machine learning (ML) community do not apply to extreme samples without careful adaptation. Besides, asymptotic results from extreme value theory (EVT) give a theoretical framework to model multivariate extreme events. Bridging these two fields, this paper details a variational autoencoder (VAE) approach for sampling multivariate heavy-tailed distributions, in which extremes of particularly large intensity are likely to occur. We illustrate the relevance of our approach on a synthetic data set and on a real data set of discharge measurements along the Danube river network. The latter shows the potential of our approach for flood risks' assessment. In addition to outperforming the vanilla VAE for the tested data sets, we also provide a comparison with a competing EVT-based generative approach. In the tested cases, our approach better captures the dependence structure between extreme events.
Keywords:
Multivariate extreme value theory
variational auto-encoders
generative models
neural network
environmental risk
Journal
J
IF:
1.2
Papers:
131
Citations:
4.1K