TITAN Meeting

Friday, Jan 23rd, 2026

cosmoGRID simulations:

Power Spectrum

Wavelet l1-norm:  sum
of wavelet coefficients
within specific amplitude
ranges across different
wavelet scales

Wavelet peaks: local maxima of wavelet
coefficient maps

N-body sims, providing DMO and baryonified full-sky κ-maps from the same IC, for 2500 cosmologies. Baryonic effects are incorporated using a shell-based Baryon Correction Model (Schneider et. al 2019).

Inference method: SBI

Training objective

\mathcal{L} = - \log p_\phi ( \theta | x )

The Scaling of Baryonic Bias with Survey Area

Determining Robust Scale Cuts

Information Content at Large Scales

Results

TARP + Ensembleing + tensiometer

Evolution of contours with survey area

BNT results

TITAN Meeting

Friday, June 27th

COLOURS

AstroStatistics Summer School

(With Hubert): PnP Mass Mapping

Wavelet ℓ1-norm theory validation (with Vilasini)

Testing impact of baryonic effects on WL HOS

Testing impact of baryonic effects on WL HOS

Results: Impact of baryons

Scale 1 (~7arcmin)

Scale 2 (~14arcmin)

Results: BNT (tomographic)

Scale 1 (~7arcmin)

Multiscale

Results: Neural Summaries (VMIM)

WST on the sphere (JAX)

TITAN Meeting

Friday, May 16th

(With Hubert): PnP Mass Mapping

Wavelet ℓ1-norm theory validation (with Vilasini)

pyro_mcmc

Pyro enables flexible and expressive deep probabilistic modeling, unifying the best of modern deep learning and Bayesian modeling

Testing impact of baryonic effects on WL HOS

Testing impact of baryonic effects on WL HOS

Results: Impact of baryons

Scale 1 (~7arcmin)

Scale 2 (~14arcmin)

BNT transform

  • BNT Transform: A method to “null” or remove contributions from unwanted redshift ranges.
  • It reorganizes the weak-lensing data so that only specific redshift ranges contribute to the signal, making it easier to analyze.
  • It focuses on isolating lensing contributions by sorting out overlapping signals.

Results: effect on statistics

Results: BNT (tomographic)

Scale 1 (~7arcmin)

Multiscale

TITAN Meeting

Friday, March 7th

Paper 

Mass mapping impact paper extension

Investigate the same thing within an ILI framework → no need for covmat, allows to investigate bias in a more proper way, allows for use of other statistics (e.g. NN-based statistics)

Project (Hubert): PnP Mass Mapping

  • Develop a new mass-mapping method based on Plug-and-Play that provides a fast reconstruction with uncertainty quantification and is trained once

  • Plug-and-Play → framework that allows to combine deep denoising algorithms with optimization algorithms to solve inverse problems

  • Some nice results for KappaTNG with a simple U-Net denoiser

Train on cosmoSLICS:

  • 25 cosmologies
  • Make the algorithm robust for different cosmologies         suitable for inference

 

 

 

 

 

 

Wavelet ℓ1-norm theory validation (with Vilasini)

  • Vilasini has a theoretical prediction for the wavelet ℓ1-norm, for any cosmology.
  • Goal: validate this theory by producing contours with this theory, and compare them to the contours obtained from the simulations.
  • Approach​
    • calculate the wavelet ℓ1-norm on cosmoSLICS maps, interpolate with GPR emulator, and produce contours
    • instead of using the emulator, use the theoretical prediction to produce contours

Euclid (HOWLS) keypaper 2

Cosmic shear higher-order statistics analysis of DR1 data with theory

  • Do an analysis of Euclid DR1 with theory-based HOS (1PDF, ℓ1-norm, Minkowski/Betti numbers,...)
  • Would like to use the wavelet ℓ1-norm emulator.
  • (also co-lead the mass-mapping brick in HOWLS)

Project: Testing impact baryonic effects on WL HOS (+BNT)

Idea - Explore two things:

  • Information content of summary statistics as a function of scale cuts
  • Testing the impact of baryonic effects on posterior contours

This will show:

  1. On what range of scales can the different statistics be used without explicit model for baryons
  2. Answer the question: how much extra information beyond the PS these statistics can access in practice

Some initial results

BNT transform

  • When we observe cosmic shear, contributions come from mass at different redshifts (distances).
  • This creates projection effects: large and small-scale structures get mixed up.
  • These effects make it harder to accurately analyze data and extract information

BNT transform

  • BNT Transform: A method to “null” or remove contributions from unwanted redshift ranges.
  • It reorganizes the weak-lensing data so that only specific redshift ranges contribute to the signal, making it easier to analyze.
  • It focuses on isolating lensing contributions by sorting out overlapping signals.

Some results

Some results

Made with Slides.com