Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
Neural Posterior Estimation (NPE) suffers from extreme overconfidence due to its log-likelihood objective and flexible density estimators, leading to poor performance on held-out data. We introduce Cross-Fitting NPE (CF-NPE), which employs $K$-fold sample splitting to decouple posterior mean estimation from higher moment estimation. CF-NPE fits $K$ mean regressors and then a density estimator to out-of-fold residuals, achieving improved held-out log-likelihood and calibration, particularly in high-dimensional, simulation-limited scenarios.
Neural posterior estimation's overconfidence problem is solved by splitting the optimization of mean and variance, yielding better out-of-sample performance.
Neural Posterior Estimation (NPE) fits an amortized conditional density estimator to a simulated training set of complete-data realizations. NPE's training objective, log-likelihood, favors posterior approximations that assign arbitrarily high density to the training points, and typical families of amortized conditional density estimators (e.g., normalizing flows, Gaussian encoders) are flexible enough to oblige. The resulting policy of extreme overconfidence translates poorly to held-out data. Early stopping and other regularizers can prevent this degeneracy, but they are limited to selecting posterior approximations along an unfavorable optimization trajectory set by in-sample residuals, where the posterior mean is underfit while the posterior variance is overfit. To avoid this trade-off, we propose Cross-Fitting NPE (CF-NPE), which uses sample splitting to separate conditional-mean estimation from the estimation of the posterior's higher moments. Using $K$-fold cross-fitting, CF-NPE fits $K$ posterior-mean regressors and then fits a conditional density estimator to the resulting out-of-fold residuals. In experiments with synthetic data and in a cosmological application, CF-NPE achieves higher held-out log-likelihood than standard NPE and typically improves calibration. CF-NPE is most useful in high-dimensional, simulation-limited settings.