REVIEW 7 cited by
Hybrid SBI or How I Learned to Stop Worrying and Learn the Likelihood
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
We propose a new framework for the analysis of current and future cosmological surveys, which combines perturbative methods (PT) on large scales with conditional simulation-based implicit inference (SBI) on small scales. This enables modeling of a wide range of statistics across all scales using only small-volume simulations, drastically reducing computational costs, and avoids the assumption of an explicit small-scale likelihood. As a proof-of-principle for this hybrid simulation-based inference (HySBI) approach, we apply it to dark matter density fields and constrain cosmological parameters using both the power spectrum and wavelet coefficients, finding promising results that significantly outperform classical PT methods. We additionally lay out a roadmap for the next steps necessary to implement HySBI on actual survey data, including consideration of bias, systematics, and customized simulations. Our approach provides a realistic way to scale SBI to future survey volumes, avoiding prohibitive computational costs.
Forward citations
Cited by 7 Pith papers
-
Modeling the Cosmological Lyman-$\alpha$ Forest at the Field Level
An effective-field-theory forward model reproduces the Lyman-alpha forest field from a hydrodynamic simulation at percent level, down to a few megaparsecs, using the same initial conditions.
-
Constraining Dynamical Dark Energy from Galaxy Clustering with Simulation-Based Priors
Adding BOSS galaxy clustering with simulation-based priors modeled as Gaussian mixtures shifts the DESI plus CMB plus supernova constraints on dark energy toward a cosmological constant and improves the w0-wa figure o...
-
PatchNet: A hierarchical approach for neural field-level inference from Quijote Simulations
Combining patch-level neural summaries with power spectrum and bispectrum extracts roughly as much cosmological information from dark matter simulations as wavelet statistics, apparently nearing the information limit ...
-
Simulation-Efficient Cosmological Inference with Multi-Fidelity SBI
A multi-fidelity SBI method using feature matching and knowledge distillation outperforms weight-initialization transfer learning at small high-fidelity simulation budgets.
-
Modeling Galaxy Surveys with Hybrid SBI
Combining perturbation theory on large scales with simulation-based inference on small scales, applied to redshift-space galaxy power spectra, yields 20% tighter constraints on Omega_m and 60% tighter on sigma_8 than ...
-
Fast Sampling of Cosmological Initial Conditions with Gaussian Neural Posterior Estimation
Cosmological initial conditions can be sampled from a learned Gaussian posterior with a Fourier-diagonal covariance, giving thousands of reconstructions in seconds on a GPU.
-
Learning Optimal and Interpretable Summary Statistics of Galaxy Catalogs with SBI
Jointly training a graph neural network with a normalizing flow yields low-dimensional summary statistics from simulated galaxy catalogs that support likelihood-free inference of Omega_m, and can be interpreted via co...
Discussion (0). Continue with ORCID to comment.