Weak-lensing Mass Reconstruction of Galaxy Clusters with Convolutional Neural Network

TL;DR

This paper presents a convolutional neural network approach for reconstructing galaxy cluster mass distributions from weak-lensing data, outperforming traditional methods and enabling model-independent mass measurements.

Contribution

The authors develop a CNN-based method trained on simulated data that improves mass reconstruction accuracy and suppresses artifacts compared to traditional techniques.

Findings

CNN outperforms traditional reconstruction methods

Improved pixel-to-pixel correlation with true mass distribution

Lifts mass-sheet degeneracy for large field reconstructions

Abstract

We introduce a novel method for reconstructing the projected matter distributions of galaxy clusters with weak-lensing (WL) data based on convolutional neural network (CNN). Training datasets are generated with ray-tracing through cosmological simulations. We control the noise level of the galaxy shear catalog such that it mimics the typical properties of the existing ground-based WL observations of galaxy clusters. We find that the mass reconstruction by our multi-layered CNN with the architecture of alternating convolution and trans-convolution filters significantly outperforms the traditional reconstruction methods. The CNN method provides better pixel-to-pixel correlations with the truth, restores more accurate positions of the mass peaks, and more efficiently suppresses artifacts near the field edges. In addition, the CNN mass reconstruction lifts the mass-sheet degeneracy when applied to our projected cluster mass estimation from sufficiently large fields. This implies that this CNN algorithm can be used to measure cluster masses in a model-independent way for future wide-field WL surveys.