Spatial Variational Auto-Encoding via Matrix-Variate Normal Distributions

May 18, 2017 ยท Entered Twilight ยท ๐Ÿ› SDM

๐ŸŒ… TWILIGHT: Old Age
Predates the code-sharing era โ€” a pioneer of its time

"Last commit was 8.0 years ago (โ‰ฅ5 year threshold)"

Evidence collected by the PWNC Scanner

Repo contents: README.md, celeba.py, cifar_reader.py, generator.py, main.py, ops.py, results, vae.py

Authors Zhengyang Wang, Hao Yuan, Shuiwang Ji arXiv ID 1705.06821 Category cs.LG: Machine Learning Cross-listed cs.CV, cs.NE, stat.ML Citations 8 Venue SDM Repository https://github.com/divelab/svae โญ 14 Last Checked 4 months ago
Abstract
The key idea of variational auto-encoders (VAEs) resembles that of traditional auto-encoder models in which spatial information is supposed to be explicitly encoded in the latent space. However, the latent variables in VAEs are vectors, which can be interpreted as multiple feature maps of size 1x1. Such representations can only convey spatial information implicitly when coupled with powerful decoders. In this work, we propose spatial VAEs that use feature maps of larger size as latent variables to explicitly capture spatial information. This is achieved by allowing the latent variables to be sampled from matrix-variate normal (MVN) distributions whose parameters are computed from the encoder network. To increase dependencies among locations on latent feature maps and reduce the number of parameters, we further propose spatial VAEs via low-rank MVN distributions. Experimental results show that the proposed spatial VAEs outperform original VAEs in capturing rich structural and spatial information.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

๐Ÿ“œ Similar Papers

In the same crypt โ€” Machine Learning