Dynamic Likelihood-free Inference via Ratio Estimation (DIRE)

By Traiko Dinev, Michael U Gutmann. 2018

27 citations recorded by the source corpus.

Parametric statistical models that are implicitly defined in terms of a stochastic data generating process are used in a wide range of scientific disciplines because they enable accurate modeling. However, learning the parameters from observed data is generally very difficult because their likelihood function is typically intractable. Likelihood-free Bayesian inference methods have been proposed which include the frameworks of approximate Bayesian computation (ABC), synthetic likelihood, and its recent generalization that performs likelihood-free inference by ratio estimation (LFIRE). A major difficulty in all these methods is choosing summary statistics that reduce the dimensionality of the data to facilitate inference. While several methods for choosing summary statistics have been proposed for ABC, the literature for synthetic likelihood and LFIRE is very thin to date. We here address this gap in the literature, focusing on the important special case of time-series models. We show that convolutional neural networks trained to predict the input parameters from the data provide suitable summary statistics for LFIRE. On a wide range of time-series models, a single neural network architecture produced equally or more accurate posteriors than alternative methods.

Subfields: Artificial Intelligence, Statistics and Probability, Probabilistic & Bayesian Methods

How researchers see this paper

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

In Artificial Intelligence

1. Gilles Louppe
2. Joeri Hermans
3. Volodimir Begy
4. Kyle Cranmer
5. J. Brehmer
6. Michael U Gutmann
7. David S. Greenberg
8. Jakob H. Macke

In Statistics and Probability

1. Gilles Louppe
2. Kyle Cranmer
3. J. Brehmer
4. Joeri Hermans
5. Volodimir Begy
6. Jukka Corander
7. Michael U Gutmann
8. Samuel Kaski

In Probabilistic & Bayesian Methods

1. Gilles Louppe
2. Joeri Hermans
3. Kyle Cranmer
4. Volodimir Begy
5. J. Brehmer
6. Michael U Gutmann
7. David S. Greenberg
8. Jakob H. Macke

Citations in the corpus

• Likelihood-free MCMC with Amortized Approximate Ratio Estimators (2019, background): “Other works investigated learning summary statistics (Fearnhead & Prangle, 2012; Dinev & Gutmann, 2018; Wong et al., 2018).”
• Mining for Dark Matter Substructure: Inferring Subhalo Population Properties from Strong Lenses with Machine Learning (2019, background): “…2016c, 2016a; Gutmann et al. 2017; Louppe & Cranmer 2017; Lueckmann et al. 2017; Papamakarios et al. 2017; Tran et al. 2017; Chen et al. 2018; Dinev & Gutmann 2018; Grathwohl et al. 2018; Huang et al. 2018; Kingma & Dhariwal 2018; Lueckmann et al. 2018; Papamakarios et al. 2018; Alsing et…”
• Automatic Posterior Transformation for Likelihood-Free Inference (2019, background): “Several recent studies investigated the use of (neural) classifiers for hypothesis testing or estimation of likelihood ratios: (Dutta et al., 2016; Gutmann et al., 2018) train binary classifiers to discriminate between the conditional and the marginal distribution, (Cran-mer et al., 2015) trains…”
• Likelihood-Free Inference by Ratio Estimation (2016, methodology): “The intrinsic connection of the proposed approach to classification facilitates the learning of summaries from raw data (Dinev and Gutmann, 2018) thereby partly addressing this point.”