Synthetic Likelihood

Synthetic likelihood is a simulation-based inference technique used to estimate parameters of complex models with intractable likelihood functions. Current research focuses on improving robustness to model misspecification, exploring various model architectures like energy-based models and neural networks to approximate the likelihood, and developing more efficient algorithms to reduce computational cost. These advancements are crucial for reliable inference in diverse fields, ranging from political science (analyzing election surveys) to neuroscience (modeling complex biological systems) and improving the safety and reliability of large language models. The ultimate goal is to enable accurate and trustworthy inference even when the underlying model is imperfect.

Papers

November 18, 2024

The Statistical Accuracy of Neural Posterior and Likelihood Estimation
David T. Frazier, Ryan Kelly, Christopher Drovandi, David J. Warne
Scientific Inference Maximum Likelihood Approximate Bayesian Computation Neural Posterior Estimation Posterior Network Neural Likelihood Synthetic Likelihood

December 23, 2023

Evaluating District-based Election Surveys with Synthetic Dirichlet Likelihood
Adway Mitra, Palash Dey
Posterior Inference Voter Model Synthetic Likelihood District Based Election

December 14, 2023

Self-Evaluation Improves Selective Generation in Large Language Models
Jie Ren, Yao Zhao, Tu Vu, Peter J. Liu, Balaji Lakshminarayanan
Large Language Model Generation Task Self Feedback Token Prediction Open Ended Generation Sequence Likelihood Synthetic Likelihood

May 25, 2023

Learning Robust Statistics for Simulation-based Inference under Model Misspecification
Daolang Huang, Ayush Bharti, Amauri Souza, Luigi Acerbi, Samuel Kaski
Simulation Based Inference Model Misspecification Neural Posterior Estimation Synthetic Likelihood

February 9, 2023

Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samples
Marco Jiralerspong, Avishek Joey Bose, Ian Gemp, Chongli Qin, Yoram Bachrach, Gauthier Gidel
Generative Model Strong Generalization External Sample Data Divergence Synthetic Likelihood

January 31, 2023

Misspecification-robust Sequential Neural Likelihood for Simulation-based Inference
Ryan P. Kelly, David J. Nott, David T. Frazier, David J. Warne, Chris Drovandi
Simulation Based Inference Model Misspecification Approximate Bayesian Computation Sequential Neural Synthetic Likelihood

October 26, 2022

Maximum Likelihood Learning of Unnormalized Models for Simulation-Based Inference
Pierre Glaser, Michael Arbel, Samo Hromadka, Arnaud Doucet, Arthur Gretton
Scientific Inference Energy Based Model Simulation Based Inference Maximum Likelihood Training Unnormalized Statistical Model Synthetic Likelihood

Synthetic Likelihood

Papers

The Statistical Accuracy of Neural Posterior and Likelihood Estimation

Evaluating District-based Election Surveys with Synthetic Dirichlet Likelihood

Self-Evaluation Improves Selective Generation in Large Language Models

Learning Robust Statistics for Simulation-based Inference under Model Misspecification

Feature Likelihood Divergence: Evaluating the Generalization of Generative Models Using Samples

Misspecification-robust Sequential Neural Likelihood for Simulation-based Inference

Maximum Likelihood Learning of Unnormalized Models for Simulation-Based Inference