Emulation of SPHEREx Galaxy Power Spectra I: Neural Network Details and Optimization

TL;DR

This paper introduces a neural network emulator that rapidly predicts SPHEREx galaxy power spectra, significantly speeding up analysis while maintaining high accuracy and agreement with theoretical models.

Contribution

The authors develop a neural network architecture combining fully-connected and transformer layers to emulate galaxy power spectra efficiently for SPHEREx data analysis.

Findings

Achieves 900x faster predictions than traditional pipelines.

Maintains median Δχ² of 0.069 compared to test data.

Emulator results agree with EFT predictions within 1σ.

Abstract

We present neural networks to generate redshift-space galaxy power spectrum multipoles for multiple tracer and redshift bins simultaneously given a set of input cosmology and galaxy bias parameters. This emulator utilizes a combination of fully-connected layers and transformer architecture to accurately predict galaxy power spectrum multipoles $900$ times faster than the SPHEREx pipeline. We quantify network performance using both $\Delta \chi^2$, and likelihood contours for simulated SPHEREx analyses, using two correlated tracer bins and two independent redshift bins. After optimizing network architecture, the loss function, and training set sampling strategy, we achieve $\operatorname{Med}\left( \Delta \chi^2\right) = 0.069$ when comparing to our testing set. At the contour-level our emulator agrees with EFT predictions over a realistic parameter range, with an average 1D best-fit shift of $0.078\sigma$ and $0.82 \%$ change in 1D error bars. These results demonstrate the feasibility of using neural-network emulators to accelerate SPHEREx redshift-space power-spectrum analyses.