no code implementations • 21 Apr 2024 • Jensen Hwa, Qingyu Zhao, Aditya Lahiri, Adnan Masood, Babak Salimi, Ehsan Adeli
We are able to enforce conditional independence of the diffusion autoencoder latent representation with respect to any protected attribute under the equalized odds constraint and show that this approach enables causal image generation with controllable latent spaces.