Variational inference offers scalable and flexible tools to tackle intractable Bayesian inference of modern statistical models like Bayesian neural networks and Gaussian processes. For largely over-parameterized models, however, the over-regularization property of the variational objective makes the application of variational inference challenging. Inspired by the literature on kernel methods, and in particular on structured approximations of distributions of random matrices, this paper proposes Walsh-Hadamard Variational Inference, which uses Walsh-Hadamard-based factorization strategies to reduce model parameterization, accelerate computations, and increase the expressiveness of the approximate posterior beyond fully factorized ones.
- Pub Date:
- November 2019
- Statistics - Machine Learning;
- Computer Science - Machine Learning
- Paper accepted at the 4th Workshop on Bayesian Deep Learning (NeurIPS 2019), Vancouver, Canada. arXiv admin note: substantial text overlap with arXiv:1905.11248