Zoom info: https://stanford.zoom.us/my/sihanyuan?pwd=QnpsUHZWWGJ2ekVYWmZVL3BmM0gzZz09

While upcoming cosmological surveys promise to map the universe with unprecedented precision, the challenge remains on how to optimally extract information from this wealth of data. We introduce Neural Quantile Estimation (NQE), a novel Simulation-Based Inference method based on conditional quantile regression, and its application to the field level inference of cosmological large scale structure. NQE autoregressively learns one-dimensional quantiles for each posterior dimension, conditioned on the observation data and previous posterior dimensions. When provided with sufficient training data, NQE converges to the Bayesian optimal posterior. In scenarios with limited training data, a post-processing step can be employed to ensure the posterior remains well-calibrated, with minimal computational overhead. Moreover, such post-processing calibration can mitigate biases stemming from model misspecification, notably those associated with inaccurate surrogate models and/or baryonic physics uncertainties.