Velocity recostruction with graph neural networks

Hideki Tanimura, Albert Bonnefous, Jia Liu, Sanmay Ganguly

2024-02-22

Abstract:In this work, we seek to improve the velocity reconstruction of clusters by using Graph Neural Networks -- a type of deep neural network designed to analyze sparse, unstructured data. In comparison to the Convolutional Neural Network (CNN) which is built for structured data such as regular grids, GNN is particularly suitable for analyzing galaxy catalogs. In our GNNs, galaxies as represented as nodes that are connected with edges. The galaxy positions and properties -- stellar mass, star formation rate, and total number of galaxies within 100~\mpc -- are combined to predict the line-of-sight velocity of the clusters. To train our networks, we use mock SDSS galaxies and clusters constructed from the Magneticum hydrodynamic simulations. Our GNNs reach a precision in reconstructed line-of-sight velocity of $\Delta v$=163 km/s, outperforming by $\approx$10\% the perturbation theory~($\Delta v$=181 km/s) or the CNN~($\Delta v$=179 km/s). The stellar mass provides additional information, improving the precision by $\approx$6\% beyond the position-only GNN, while other properties add little information. Our GNNs remain capable of reconstructing the velocity field when redshift-space distortion is included, with $\Delta v$=210 km/s which is again 10\% better than CNN with RSD. Finally, we find that even with an impressive, nearly 70\% increase in galaxy number density from SDSS to DESI, our GNNs only show an underwhelming 2\% improvement, in line with previous works using other methods. Our work demonstrates that, while the efficiency in velocity reconstruction may have plateaued already at SDSS number density, further improvements are still hopeful with new reconstruction models such as the GNNs studied here.

Cosmology and Nongalactic Astrophysics

What problem does this paper attempt to address?

The paper aims to improve the accuracy of reconstructing the line-of-sight (LOS) velocity of galaxy clusters by using Graph Neural Networks (GNN). Specifically, the paper addresses the following issues:

1. **Improving Velocity Reconstruction Methods**: Compared to traditional Convolutional Neural Networks (CNN), GNN is more suitable for handling sparse, unstructured data, such as galaxy catalogs. GNN can better capture global patterns and integrate various galaxy attribute information.

2. **Enhancing Accuracy**: GNN performs excellently in reconstructing the radial velocity of galaxy clusters, achieving an accuracy of ∆3=163 km/s, which is about 10% better than perturbation theory (∆3=181 km/s) and CNN (∆3=179 km/s).

3. **Handling Redshift Space Distortion (RSD)**: Even when considering redshift space distortion, GNN's accuracy can still reach ∆3=210 km/s, which is about 10% higher than CNN.

4. **Evaluating the Impact of Different Features**: The paper investigates the impact of different galaxy attributes (such as stellar mass, star formation rate, etc.) on the accuracy of velocity reconstruction. The results show that adding stellar mass information can improve accuracy by about 6%, while other attributes contribute little.

5. **Comparing the Impact of Different Galaxy Densities**: The increase in galaxy density from SDSS to DESI does not significantly improve the accuracy of velocity reconstruction, indicating that further improvements may require new models rather than simply increasing the amount of data.

In summary, the paper enhances the accuracy of galaxy cluster radial velocity reconstruction by introducing GNN and demonstrates its advantages in handling sparse, unstructured data.