Transient Detection Modelling for Gravitational-wave Optical Transient Observer (GOTO) Sky Survey
Transient Detection Modelling for Gravitational-wave Optical Transient Observer (GOTO) Sky Survey
复制标题
DOI：
10.1145/3195106.3195153
复制
发表时间：
2018-02
期刊：
影响因子：
--
通讯作者：
Aireen B. Tabacolde;Tossapon Boongoen;Natthakan Iam-on;J. Mullaney;U. Sawangwit;K. Ulaczyk
中科院分区：
文献类型：
--
作者：
Aireen B. Tabacolde;Tossapon Boongoen;Natthakan Iam-on;J. Mullaney;U. Sawangwit;K. Ulaczyk
鉴于数据采集和望远镜技术的进步，天文学近年来加入了大数据和人工智能的全球趋势。GOTO的目标是识别引力波探测的光学对应物。这需要每晚获取许多天空图像，然后对这些图像进行系统处理和分析，以提供4000万个观测来源。然后，将这些源与参考集进行比较，从而可以提取新的亮源并使用它们来形成一组对应的候选对象。大多数候选者不会代表真实的情况，因为他们检测到的亮度变化是由于数据收集和/或预处理中的错误造成的。为此，必须从假阳性中正确筛选出少数真正的候选者，以使天文学家能够有效地利用后续观测来验证他们的真实性。上述问题恰好属于数据分类，在数据分类中，候选人的多个物理测量被解释为独立变量，专家给出的标签为类别变量。这项研究旨在探索分析这一特定数据集的常规技术，从数据准备到模型开发和评估。我们的研究结果不仅为未来的发展提供了基线，而且还提供了对数据特征的彻底审查。它还将被证明在塑造获取和存储数据的方法方面对GoTO项目很有用。
Given the advancement of data acquisition and telescope technology, astronomy has joined the global trend of big data and artificial intelligence in recent years. The objective of GOTO is to identify optical counterparts to gravitational wave detections. This requires obtaining many images of the sky every night, which are then systematically processes and analysed to deliver 40-million observed sources. These sources are then compared against a reference set such that new bright sources can be extracted and used to form a set of counterpart candidates. Most of the candidates will not represent real cases, with their detected changes in brightness caused by errors in data collection and/or pre-processing. To this end, the handful of real candidates has to be correctly sifted from the false-positives to allow astronomers to effectively employ follow-up observations to verify their truth. The aforementioned problem falls nicely into data classification, where multiple physical measurements of candidates are explicated as independent variables with labels given by experts as the class variable. This research is set to explore conventional techniques to analyze this specific dataset, from data preparation through to model development and evaluation. The outcome of our research not only provides a baseline for future developments, but also pro-vides a thorough review of data characteristics. It will be also proving useful for the GOTO project in terms of shaping the approach to acquire and store data.