Martingale Posterior
A core concept in the predictive Bayesian framework that generates posterior samples directly through exchangeable predictive rules without specifying a joint likelihood. It exploits martingale properties to ensure probabilistic calibration of the posterior, offering an alternative foundation for Bayesian inference.
Timeline (last 90 days)
This study proposes a model-free, noise-aware Bayesian clustering framework that, from a predictive Bayesian perspective, constructs a martingale posterior over clustering structures and quantifies clustering uncertainty without explicitly specifying a model.
The martingale posterior uses exchangeable predictive rules to directly generate posterior samples of clustering structures; as long as the predictive rules satisfy consistency conditions, probabilistic calibration of the posterior is guaranteed without requiring parametric assumptions about the data distribution.
All Facts (2)
This study proposes a model-free, noise-aware Bayesian clustering framework that, from a predictive Bayesian perspective, constructs a martingale posterior over clustering structures and quantifies clustering uncertainty without explicitly specifying a model.
50%UnverifiedThe martingale posterior uses exchangeable predictive rules to directly generate posterior samples of clustering structures; as long as the predictive rules satisfy consistency conditions, probabilistic calibration of the posterior is guaranteed without requiring parametric assumptions about the data distribution.
50%