Explicit Density Approximation for Neural Implicit Samplers Using a Bernstein-Based Convex Divergence
Date:
Spotlight oral presentation at AISTATS 2026, given in Oral Session 6: Generative Modeling (Diffusion / Flow / Temporal / Graph). The talk presented dual-ISL, a rank-based likelihood-free objective that induces convex optimization over model densities, admits an explicit Bernstein-based density approximation, and extends to multivariate distributions through slicing.
Official AISTATS spotlight page · Conference schedule · Paper · Publications