.png){kind=icon}
.png){kind=icon}
Return
MAPLE-VAE: MAP-based LaplacE VAE for robust density estimation
.png){kind=icon}
PRE
AI
DOI：10.1016/j.patcog.2026.113284.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
• This is the first study to delineate the applicable scenarios and conditions for Gaussian, Laplace, and Student-t decoders. • The Laplace VAE is proposed to mitigate training instability in scenarios in which the feature space includes numerous zero-variance features. • A regularization term is introduced by applying a gamma prior to the shape parameter of the Laplace decoder to address feature redundancy in high-dimensional data. • The MAPLE-VAE is extended to the Cauchy distribution and applied to highly imbalanced datasets, achieving promising results.
Keywords:
Laplace VAE
Student-t decoder
Gaussian decoder
regularization
density estimation
Journal
IF:
7.6
Papers:
1.3W
Citations:
4.5W