Implementation of DDPM based oversampling method. Capable of handling tabular datasets with categorical and numerical attributes. The fitted DDPM model can be used to resample unbalanced training sets.
The datasets used by our evaluation are not included in this repository but are linked to in dataloader.py. At the time of writing, all the datasets are publicly available.