Overview

GOLCONDA (Generative modeling of convergence maps based on predicted one-point statistics) is an efficient emulator for generating weak lensing convergence ($\kappa$) maps that capture non-Gaussian information beyond traditional power spectra. It synthesizes maps directly from input cosmological statistics, eliminating reliance on computationally expensive N-body simulations.

Key Features

• 🎯 Input Flexibility: Accepts theoretical predictions or precomputed:

  • Power spectrum (two-point statistics)

  • Wavelet $\ell_1$-norm (higher-order statistics)

• ⚙️ Core Method: Iterative optimization of wavelet coefficients to match:

  • Target power spectrum

  • Marginal distributions and inter-scale dependencies

• 📊 High-Fidelity Output: Generates $\kappa$ maps preserving:

  • Input power spectra

  • Higher-order statistical properties

  • Non-Gaussian information from nonlinear structure formation

Advantages

• ⚡ faster than traditional simulations

• 💻 Reduced computational requirements

• 🔍 Accurate reproduction of cosmological statistics

• 🚀 Scalable for large-scale survey analyses (LSST, Euclid)

GitHub Repository

You can find the GitHub repository here.

Use Cases

Cosmological parameter inference with non-Gaussian statistics

Pipeline validation for Stage-IV lensing surveys

Fast generation of training data for machine learning models

Covariance matrix estimation beyond power spectra

Contact

Please feel free to reach out to tsvilasini97@gmail.com in case of any bugs or questions regarding usage.

How to Install

Clone the repo and run:

pip install .