Truncated Variational Sampling for "Black Box" Optimization of Generative Models
This work addresses the problem of optimizing generative models for researchers in machine learning, offering a more efficient and general-purpose method, though it appears incremental as it builds on existing variational approaches.
The paper tackles the optimization of generative models with binary latent variables by introducing a novel variational EM approach that uses latent states as variational parameters and employs efficient sampling procedures. The method is applied to Binary Sparse Coding and Sigmoid Belief Networks, with numerical experiments showing it effectively increases a variational free energy objective without additional analytical steps.
We investigate the optimization of two probabilistic generative models with binary latent variables using a novel variational EM approach. The approach distinguishes itself from previous variational approaches by using latent states as variational parameters. Here we use efficient and general purpose sampling procedures to vary the latent states, and investigate the "black box" applicability of the resulting optimization procedure. For general purpose applicability, samples are drawn from approximate marginal distributions of the considered generative model as well as from the model's prior distribution. As such, variational sampling is defined in a generic form, and is directly executable for a given model. As a proof of concept, we then apply the novel procedure (A) to Binary Sparse Coding (a model with continuous observables), and (B) to basic Sigmoid Belief Networks (which are models with binary observables). Numerical experiments verify that the investigated approach efficiently as well as effectively increases a variational free energy objective without requiring any additional analytical steps.