Categorical Reparameterization with Gumbel-Softmax

使用Gumble-Softmax实现离散类别隐变量的重参数化.

Deep Feature Consistent Variational Autoencoder

DFCVAE：使用特征感知损失约束深度特征一致性.

Tighter Variational Bounds are Not Necessarily Better

MIWAE：紧凑的变分下界阻碍推理网络训练.

Importance Weighted Autoencoders

IWAE：重要性加权自编码器.

Single-Shot Multi-Person 3D Pose Estimation From Monocular RGB

使用遮挡鲁棒姿态图从单目相机中重构三维姿态.

Isolating Sources of Disentanglement in Variational Autoencoders

β-TCVAE: 分离VAE解耦源中的全相关项.