O. Deniz Akyildiz

We propose a novel estimation method for latent variable models.

We propose a training-free framework for sampling by exploiting multiscale dynamics and general diffusion paths.