Projects
Variational Autoencoder (VAE)
A tiny variational autoencoder (VAE) model to encode images into latent space for latent diffusion models (LDM) in crunch computing power.
Latent Diffusion Model (LDM)
A simplified, home-made diffusion model that can be trained on 5GB VRAM
RiVAE: Riemannian Geometric Generation via Epistemic Graphs
A novel generative model that leverages Riemannian geometry and epistemic uncertainty to produce high-quality data samples by navigating learned latent manifolds.