LGITNov 21, 2020

Use of Student's t-Distribution for the Latent Layer in a Coupled Variational Autoencoder

arXiv:2011.10879v12 citations
AI Analysis

This work provides an incremental improvement in image generation accuracy and robustness for researchers working with VAEs and image synthesis.

This paper introduces a Coupled Variational Autoencoder that improves the accuracy and robustness of generated MNIST numeral replicas. It achieves this by using a Student's t-distribution for the latent layer and a coupled logarithm in the loss function.

A Coupled Variational Autoencoder, which incorporates both a generalized loss function and latent layer distribution, shows improvement in the accuracy and robustness of generated replicas of MNIST numerals. The latent layer uses a Student's t-distribution to incorporate heavy-tail decay. The loss function uses a coupled logarithm, which increases the penalty on images with outlier likelihood. The generalized mean of the generated image's likelihood is used to measure the performance of the algorithm's decisiveness, accuracy, and robustness.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes