Lossless, scalable implicit likelihood inference for cosmological fields

Abstract

We present a comparison of simulation-based inference to full, field-based analytical inference in cosmological data analysis. To do so, we explore parameter inference for two cases where the information content is calculable analytically: Gaussian random fields whose covariance depends on parameters through the power spectrum; and correlated lognormal fields with cosmological power spectra. We compare two inference techniques: i) explicit field-level inference using the known likelihood and ii) implicit likelihood inference with maximally informative summary statistics compressed via Information Maximising Neural Networks (IMNNs). We find that a) summaries obtained from convolutional neural network compression do not lose information and therefore saturate the known field information content, both for the Gaussian covariance and the lognormal cases, b) simulation-based inference using these maximally informative nonlinear summaries recovers nearly losslessly the exact posteriors of field-level inference, bypassing the need to evaluate expensive likelihoods or invert covariance matrices, and c) even for this simple example, implicit, simulation-based likelihood incurs a much smaller computational cost than inference with an explicit likelihood. This work uses a new IMNN implementation in Jax that can take advantage of fully-differentiable simulation and inference pipeline. We also demonstrate that a single retraining of the IMNN summaries effectively achieves the theoretically maximal information, enhancing the robustness to the choice of fiducial model where the IMNN is trained.

A Full $w$CDM Analysis of KiDS-1000 Weak Lensing Maps using Deep Learning

“…[23,24]) or machine learning based methods (e.g. [25][26][27][28][29]). These higher order statistics have the potential to greatly improve the cosmological parameter constraints and reduce the systematic effects.…”

Fisher Forecasts for Primordial non-Gaussianity from Persistent Homology

“…The tradeoff is that the likelihood for our persistence diagrams is only implicitly defined. Parameter estimation in the context of implicit likelihoods is precisely within the purview of the rapidly advancing field of simulation-based inference ( [80][81][82][83], see [84] for a recent review and [85][86][87][88][89] for cosmological applications). Within simulation-based inference it has been advocated (see e.g.…”

Snowmass White Paper: Cosmology at the Theory Frontier

“…More generally, it was argued in [246] that a map-level analysis may significantly improve the constraining power over previous forecasts. Thus, ML approaches such as simulation-based inference involving the forward modeling of large-scale structure maps [360][361][362][363][364][365][366][367][368][369] may have the potential to dramatically impact the search for primordial non-Gaussianity.…”