An Improved Variational Approximate Posterior for the Deep Wishart Process - Oral Session 1

Explore a conference talk from the Uncertainty in Artificial Intelligence (UAI) 2023 Oral Session 1 that delves into an improved variational approximate posterior for the Deep Wishart Process. Learn about deep kernel processes, a class of deep Bayesian models that combine the flexibility of neural networks with Gram matrix operations. Discover how the Deep Wishart Process (DWP) relates to deep Gaussian processes (DGP) and offers rotational symmetry invariance. Examine the limitations of previous variational approximate posterior methods and understand the proposed enhancement using linear combinations of rows and columns in the Bartlett decomposition. Gain insights into how this improvement leads to better predictive performance without significant additional computational cost. Access the presentation slides to visualize key concepts and findings from this 24-minute talk by Sebastian W. Ober and colleagues.