Block Neural Autoregressive Flow

  • 2019-04-09 13:54:55
  • Nicola De Cao, Ivan Titov, Wilker Aziz
  • 46

Abstract

Normalising flows (NFS) map two density functions via a differentiablebijection whose Jacobian determinant can be computed efficiently. Recently, asan alternative to hand-crafted bijections, Huang et al. (2018) proposed neuralautoregressive flow (NAF) which is a universal approximator for densityfunctions. Their flow is a neural network (NN) whose parameters are predictedby another NN. The latter grows quadratically with the size of the former andthus an efficient technique for parametrization is needed. We propose blockneural autoregressive flow (B-NAF), a much more compact universal approximatorof density functions, where we model a bijection directly using a singlefeed-forward network. Invertibility is ensured by carefully designing eachaffine transformation with block matrices that make the flow autoregressive and(strictly) monotone. We compare B-NAF to NAF and other established flows ondensity estimation and approximate inference for latent variable models. Ourproposed flow is competitive across datasets while using orders of magnitudefewer parameters.

 

Introduction (beta)

None

 

Conclusion (beta)

None