PolyMNIST Dataset

class multivae.data.datasets.MMNISTDataset(data_path, transform=None, target_transform=None, split='train', download=False, missing_ratio=0, keep_incomplete=True)

Multimodal PolyMNIST Dataset from ‘Generalized Multimodal Elbo’ Sutter et al 2021.

This dataset class has a parameter ‘missing_ratio’ that allows to simulate a dataset with missing values (Missing At Random).

>>> from multivae.data.datasets import MMNISTDataset
>>> dataset = MMNISTDataset(
...            data_path = 'your_data_path',
...            split = 'train',
...            download = True, #to download the dataset
...            missing_ratio = 0.2 # 20% of missing data
...        )