Publications

C3NN-SBI: Learning Hierarchies of N-Point Statistics from Cosmological Fields with Physics-Informed Neural Networks

Authors: K. Lehman, Z. Gong, D. Gebauer, S. Seitz, J. Weller

arXiv:2602.16768 • 2026

A simulation-based inference framework using a constrained C3NN architecture to learn summary statistics that correspond to N-point correlation functions of specified orders from cosmological fields.

SBi3PCF: Simulation-based inference with the integrated 3PCF

Authors: D. Gebauer, A. Halder, S. Seitz, D. Anbajagane

JCAP 06(2026)036 • 2026

A framework for higher-order weak lensing analysis using the integrated 3-point correlation function, employing masked autoregressive flows for neural likelihood estimation. Including the i3PCF yields a 63.8% improvement in the figure of merit for cosmological parameters.

Cosmology with second and third-order shear statistics for the Dark Energy Survey: Methods and simulated analysis

Authors: R. C. H. Gomes, S. Sugiyama, B. Jain, M. Jarvis, D. Anbajagane, M. Gatti, D. Gebauer, Z. Gong, A. Halder, G. A. Marques, S. Pandey, J. L. Marshall, DES Collaboration

Phys. Rev. D 112, 123514 • 2025

A pipeline for robust inference of cosmological parameters using second- and third-order shear statistics, demonstrating an 83% improvement in the figure of merit when combining third-order with second-order statistics for DES Year 3 data.

C3NN: Cosmological Correlator Convolutional Neural Network -- an interpretable machine learning tool for cosmological analyses

Authors: Z. Gong, A. Halder, A. Bohrdt, S. Seitz, D. Gebauer

ApJ 971 156 • 2024

A convolutional neural network architecture whose outputs can be expressed in terms of analytically tractable N-point correlation functions, enabling interpretable and physically meaningful feature extraction from cosmological fields.

In Preparation

DES Year 3: Cosmology with the Integrated 3-point Correlation Function of cosmic shear

In preparation

A. Halder, Z. Gong, D. Gebauer, S. Seitz, DES Collaboration

Measurement and cosmological analysis of the integrated 3-point correlation function of DES Y3 cosmic shear, combined with the shear 2PCF, using analytical modelling based on the response function approach to perturbation theory with MOPED compression and emulator-accelerated inference.

SBi3PCF II: Cosmology from DES Y3 cosmic shear

In preparation

D. Gebauer, A. Halder, C. Uhlemann, Z. Gong

A fully blinded application of the SBi3PCF simulation-based inference framework to DES Y3 data, jointly analysing the cosmic shear 2-point and integrated 3-point correlation functions with forward-modelled CosmoGridV1 mocks. The joint wCDM analysis yields S8 = 0.780 (+0.026, −0.023) and improves the (Ωm, S8, w0) figure of merit by 58% over the 2PCF alone.

Where the matter PDF keeps its cosmological information: an interpretable simulation-based inference study

In preparation

D. Gebauer, C. Uhlemann

A systematic study of where the smoothed matter PDF keeps its cosmological information: cumulant- and quantile-based compressions of the Quijote simulations are compared across smoothing scales and redshifts within an identical neural inference pipeline, and SHAP attributions localise the signal in cosmic environments and guide the design of improved summary statistics.