DeepZipper II：利用深度学习在暗能量巡天数据中搜索透镜超新星
宇宙学与河外天体物理
2023-02-01 v2
摘要
引力透镜超新星(LSNe)是宇宙膨胀的重要探针，但它们仍然罕见且难以发现。当前的宇宙学巡天总共可能包含5-10个LSNe，而下一代实验预计将包含数百到数千个此类系统。我们在已观测的暗能量巡天(DES)五年超新星场——10个约3平方度的天区，在五年内大约每六晚以波段成像——中搜索这些系统。为执行搜索，我们采用DeepZipper方法：一种在多分支深度学习架构上训练、基于LSNe图像级模拟的模型，可从图像时间序列中同时学习空间与时间关系。我们发现，我们的方法在DES超新星场数据上获得了61.13%的LSN召回率和0.02%的假阳性率。DeepZipper从一个限制星等( 22.5)的包含3,459,186个系统的星表中选出了2,245个候选体。我们采用人工视觉检查来审查网络选出的系统，并在DES超新星场中发现了三个LSNe候选体。
引用
@article{arxiv.2204.05924,
title = {DeepZipper II: Searching for Lensed Supernovae in Dark Energy Survey Data with Deep Learning},
author = {Robert Morgan and B. Nord and K. Bechtol and A. Möller and W. G. Hartley and S. Birrer and S. J. González and M. Martinez and R. A. Gruendl and E. J. Buckley-Geer and A. J. Shajib and A. Carnero Rosell and C. Lidman and T. Collett and T. M. C. Abbott and M. Aguena and F. Andrade-Oliveira and J. Annis and D. Bacon and S. Bocquet and D. Brooks and D. L. Burke and M. Carrasco Kind and J. Carretero and F. J. Castander and C. Conselice and L. N. da Costa and M. Costanzi and J. De Vicente and S. Desai and P. Doel and S. Everett and I. Ferrero and B. Flaugher and D. Friedel and J. Frieman and J. García-Bellido and E. Gaztanaga and D. Gruen and G. Gutierrez and S. R. Hinton and D. L. Hollowood and K. Honscheid and K. Kuehn and N. Kuropatkin and O. Lahav and M. Lima and F. Menanteau and R. Miquel and A. Palmese and F. Paz-Chinchón and M. E. S. Pereira and A. Pieres and A. A. Plazas Malagón and J. Prat and M. Rodriguez-Monroy and A. K. Romer and A. Roodman and E. Sanchez and V. Scarpine and I. Sevilla-Noarbe and M. Smith and E. Suchyta and M. E. C. Swanson and G. Tarle and D. Thomas and T. N. Varga},
journal= {arXiv preprint arXiv:2204.05924},
year = {2023}
}
备注
Accepted by ApJ