Open braceal opened 4 years ago
Current CVAE uses ConvTranspose2d, may lead to checkerboard artifacts. Consider using upsampling modules instead. https://distill.pub/2016/deconv-checkerboard/
Sparsity https://arxiv.org/pdf/1812.07238.pdf
Arbitrary posterior distribution for vae: https://arxiv.org/pdf/1702.07904.pdf
Variational Autoencoders with Inverse Autoregressive Flows: Allows modelling multi-modal latent distributions as could be present in contact matrix datasets. https://bjlkeng.github.io/posts/variational-autoencoders-with-inverse-autoregressive-flows/