###### tags: `Papers`
# EWAF 2023
here is an overview of the changes:
## Previous submission
* a case study on new adult datasets
* model deployment under distribution shift.
* local vs global models
## new draft
* data augmentation using clustering based on MMD measure (proposeing a prerpocessing intervention as contribution)
* measure data bias before and after augmenting (without model deployment and evaluation)
* A clustering framework with automatic identification of optimal number of clusters based on context similarity
* one of the qualitatvie results is shown below: first image shows original racial distribution of 3 groups for 5 states. second shows balanced group distributions for the same states after the augmentation