Variational Inference with Orthogonal Normalizing Flows

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Sorting out Lipschitz function approximation (2018, background): “The Björck orthonormalization can be shown to always converge as long as the condition ||WW − I||2 < 1 is satisfied (Hasenclever et al., 2017).”
Deep Generative Modelling: A Comparative Review of VAEs, GANs, Normalizing Flows, Energy-Based and Autoregressive Models (2021, background): “A higher rank generalisation of the matrix determinant lemma has been applied to planar flows, known as Sylvester flows, removing the severe bottleneck thus allowing greater representation ability [13], [78].”
The Expressive Power of a Class of Normalizing Flow Models (2020, methodology): “Several flows parameterized by matrices were inspired by results from linear algebra and thus enjoy this property (Tomczak and Welling, 2016; Hasenclever et al., 2017; Ho et al., 2019; Berg et al., 2018), where the last one is a matrix-form generalization of the planar flow.”
Normalizing Flows: An Introduction and Review of Current Methods (2020, background): “Some sufﬁcient conditions for the invertibility of Sylvester transformations are discussed in Hasenclever et al. [38] and van den Berg et al. [97].”