Neural Approximate Sufficient Statistics for Implicit Models
This addresses a fundamental challenge in statistical inference for implicit models, offering a novel approach that enhances existing methods without requiring density estimation.
The paper tackles the problem of automatically constructing summary statistics for implicit generative models with intractable likelihoods by framing it as learning mutual information maximizing representations using deep neural networks, which boosts performance on a range of tasks when applied to approximate Bayesian computation and neural likelihood methods.
We consider the fundamental problem of how to automatically construct summary statistics for implicit generative models where the evaluation of the likelihood function is intractable, but sampling data from the model is possible. The idea is to frame the task of constructing sufficient statistics as learning mutual information maximizing representations of the data with the help of deep neural networks. The infomax learning procedure does not need to estimate any density or density ratio. We apply our approach to both traditional approximate Bayesian computation and recent neural likelihood methods, boosting their performance on a range of tasks.