This is the official implementation of "A Geometric Perspective on Variational Autoencoders" (NeurIPS 2022)

This code uses a version of python3.6.

Note: The method should be soon added to pythae.

To install requirement run

pip install -r requirements.txt

Data folders

The data must be located in data_folders:

MNIST

The provided code requires a file mnist_32x32.npz to be located in data_folders/mnist/. The data must be in the range [0, 255] and loadable as follows:

    import numpy as np
    mnist_digits = np.load(args.path_to_train)
    train_data = mnist_digits['x_train'] # data of shape 60000x32x32x1 in [0-255]
    train_targets = mnist_digits['y_train'] # corresponding labels

In data_folders/mnist/test_folder must be located 10k test images in .png format used for metric computation

CIFAR10

The provided code requires a file cifar_10.npz to be located in data_folders/cifar/. The data must be in the range of [0, 255] and lodable as follows:

    import numpy as np
    cifar_data = np.load(args.path_to_train)
    train_data = cifar_data['x_train'] # data of shape 50000x32x32x3 in [0-255]
    train_targets = cifar_data['y_train'] # corresponding labels

In data_folders/cifar/test_folder must be located 10k test images in .png format used for metric computation

Celeba

The provided code requires a file train_data.pt to be located in data_folders/celeba/. The data must be a big tensor of shape n_samplesx3x64x64 in the range [0, 1] and loadable as follows:

    import torch
    train_data = torch.load(os.path.join(args.path_to_train, 'train_data.pt')) # data of shape 162770x64x64x3 in the range of [0-1]
    val_data = torch.load(os.path.join(args.path_to_train, 'val_data.pt')) # data of shape 19867x64x64x3 in the range of [0-1]

In data_folders/celeba/test/test must be located the test images in .png format used for metric computation

SVHN

The provided code requires a file train_32x32.mat to be located in data_folders/svhn/. The data must be in the rnage [0, 255] and loadable as follows:

    from scipy.io import loadmat
    svnh_digits = loadmat(args.path_to_train)['X'] # data of shape 32x32x3x73257 in the range of [0-255]
    svnh_targets = loadmat(args.path_to_train)['y'] # corresponding labels

In data_folders/svhn/test_folder must be located the test images in .png format used for metric computation.

OASIS

The provided code requires a file OASIS.npz to be located in data_folders/oasis/. The data must be in the range of [0, 255] and you must ensure that each data image has a maximum voxel value of 255 and a minimum of 0. The data must be loadable as follows

    import numpy as np
    oasis_data = np.load(args.path_to_train)
    train_data = oasis_data['x_train'] # data of shape 416x208x176x1 in the range of [0-255]
    train_targets = torch.tensor(oasis_data['y_train'] # corresponding targets

Performing experiments

The commandines to train a model, generate new data and compute the metrics are available in models_to_train.sh.

Reference

@article{chadebec2022geometric,
  title={A geometric perspective on variational autoencoders},
  author={Chadebec, Cl{\'e}ment and Allassonni{\`e}re, St{\'e}phanie},
  journal={Advances in Neural Information Processing Systems},
  volume={35},
  pages={19618--19630},
  year={2022}
}