arXiv subjects

Yiyan Yang

Publications and source records attributed to Yiyan Yang.

2 recordsLinked to original sources

arXiv2022

Distribution of fast radio burst dispersion measures in CHIME/FRB Catalog 1: implications on the origin of FRBs

Recently, CHIME/FRB project published its first fast radio burst (FRB) catalog (hereafter, Catalog 1), which totally contains 536 unique bursts. With the help of the latest set of FRBs in this large-size catalog, we aim to investigate the dispersion measure (DM) or redshift ($z$) distribution of the FRB population, and solution of this problem could be used to clarify the question of FRB origin. In this study, we adopted the M\&E 2018 model, to fit the observed $z$ distribution of FRBs in Catalog 1. In the M\&E 2018 model, we are mostly interested in the $Φ(z)$ function, i.e., number of bursts per proper time per comoving volume, which is represented by the star formation rate (SFR) with a power-law index $n$. Our estimated value of $n$ is $0.0_{-0.0}^{+0.6}$ ($0.0_{-0.0}^{+2.1}$) at the 68 (95) per cent confidence level, implying that the FRB population evolves with redshift consistent with, or faster than, the SFR. Specially, the consistency of the $n$ values estimated by this study and the SFR provides a potential support for the hypothesis of FRBs originating from young magnetars.

astro-ph.HE↗

arXiv2020

Learning Local Neighboring Structure for Robust 3D Shape Representation

Mesh is a powerful data structure for 3D shapes. Representation learning for 3D meshes is important in many computer vision and graphics applications. The recent success of convolutional neural networks (CNNs) for structured data (e.g., images) suggests the value of adapting insight from CNN for 3D shapes. However, 3D shape data are irregular since each node's neighbors are unordered. Various graph neural networks for 3D shapes have been developed with isotropic filters or predefined local coordinate systems to overcome the node inconsistency on graphs. However, isotropic filters or predefined local coordinate systems limit the representation power. In this paper, we propose a local structure-aware anisotropic convolutional operation (LSA-Conv) that learns adaptive weighting matrices for each node according to the local neighboring structure and performs shared anisotropic filters. In fact, the learnable weighting matrix is similar to the attention matrix in the random synthesizer -- a new Transformer model for natural language processing (NLP). Comprehensive experiments demonstrate that our model produces significant improvement in 3D shape reconstruction compared to state-of-the-art methods.

cs.CV↗