Supporting code to the paper: Variational autoencoders in the presence of low-dimensional data: landscape and implicit bias
Experiments related to the linear dataset is in the seed_linpadding_expts.sh file;
experiments related to the sigmoid dataset is in the sigmoid_vad_padding_expts.sh file;
experiments related to the sphere dataset is in the sphere_vad_padding_expts.sh file;
Required packages include:
- JAX
- Flax
- NumPy
- OpenCV-Python