Likelihood-free inference with an improved cross-entropy estimator
This is an incremental improvement for researchers in statistical inference and machine learning.
The paper tackles the problem of likelihood-free inference by extending prior work that uses neural networks as surrogate models, showing how augmented training data can be used to create a new cross-entropy estimator that improves sample efficiency compared to previous loss functions.
We extend recent work (Brehmer, et. al., 2018) that use neural networks as surrogate models for likelihood-free inference. As in the previous work, we exploit the fact that the joint likelihood ratio and joint score, conditioned on both observed and latent variables, can often be extracted from an implicit generative model or simulator to augment the training data for these surrogate models. We show how this augmented training data can be used to provide a new cross-entropy estimator, which provides improved sample efficiency compared to previous loss functions exploiting this augmented training data.