兹威基瞬变设施的机器学习
天体物理仪器与方法
2019-02-07 v1
摘要
兹威基瞬变设施是一项多滤光片的大型光学巡天，每夜产生数十万条瞬变警报。我们在此描述各种机器学习（ML）实现与计划，以通过利用数据的时序性质并进一步将其与其他数据集结合，来最大化利用这一大型数据集。我们从区分伪候选与真实候选、区分恒星与星系的初步步骤开始，进而将真实天体分类为各类。除常规方法（例如基于从光变曲线提取的特征）外，我们还描述了替代方法的早期计划，包括使用域适应和深度学习。类似地我们描述了探测快速移动小行星的努力。我们还描述了利用 Zooniverse 平台通过创建训练样本和主动学习来辅助分类。最后我们从 ML 视角提及 ZTF 与 LSST 的协同方面。
引用
@article{arxiv.1902.01936,
title = {Machine Learning for the Zwicky Transient Facility},
author = {Ashish Mahabal and Umaa Rebbapragada and Richard Walters and Frank J. Masci and Nadejda Blagorodnova and Jan van Roestel and Quan-Zhi Ye and Rahul Biswas and Kevin Burdge and Chan-Kao Chang and Dmitry A. Duev and V. Zach Golkhou and Adam A. Miller and Jakob Nordin and Charlotte Ward and Scott Adams and Eric C. Bellm and Doug Branton and Brian Bue and Chris Cannella and Andrew Connolly and Richard Dekany and Ulrich Feindt and Tiara Hung and Lucy Fortson and Sara Frederick and C. Fremling and Suvi Gezari and Matthew Graham and Steven Groom and Mansi M. Kasliwal and Shrinivas Kulkarni and Thomas Kupfer and Hsing Wen Lin and Chris Lintott and Ragnhild Lunnan and John Parejko and Thomas A. Prince and Reed Riddle and Ben Rusholme and Nicholas Saunders and Nima Sedaghat and David L. Shupe and Leo P. Singer and Maayane T. Soumagnac and Paula Szkody and Yutaro Tachibana and Kushal Tirumala and Sjoert van Velzen and Darryl Wright},
journal= {arXiv preprint arXiv:1902.01936},
year = {2019}
}
备注
Published in PASP Focus Issue on the Zwicky Transient Facility (doi: 10.1088/1538-3873/aaf3fa). 14 Pages, 8 Figures