Star cluster classification in the PHANGS– HST survey: Comparison between human and machine learning approaches
Star cluster classification in the PHANGS– HST survey: Comparison between human and machine learning approaches
复制标题
PHANGS™ HST 调查中的星团分类：人类和机器学习方法的比较
DOI：
10.1093/mnras/stab2087
复制
发表时间：
2021
影响因子：
4.8
通讯作者：
Chevance, Mélanie
中科院分区：
文献类型：
--
作者：
Whitmore, Bradley C;Lee, Janice C;Chandar, Rupali;Thilker, David A;Hannon, Stephen;Wei, Wei;Huerta, E A;Bigiel, Frank;Boquien, Médéric;Chevance, Mélanie
完成后，PHANGS-HST项目将提供大约50万个致密星团和星团的普查，以及其中大约20万个天体的人类形态分类。这些大量数据促使人们开发出一种更客观、可重复的方法来帮助执行源分类。在本文中，我们考虑了五个 PHANGS-HST 星系（NGC 628、NGC 1433、NGC 1566、NGC 3351、NGC 3627）的结果，使用了使用深度迁移学习技术训练的两种卷积神经网络架构（RESNET 和 VGG）的分类。结果与人类执行的分类进行比较。主要结果是，神经网络分类在质量上与人类分类相当，1 类簇（对称、集中集中）的典型一致性约为 70% 至 80%，2 类簇（不对称、集中集中）的一致性约为 40% 至 70%。如果将 1 类和 2 类一起考虑，则一致性为 82% ± 3%。检查对幅度、拥挤度和背景表面亮度的依赖性。其中包括对人类分类所用标准和方法的详细描述，以及对 PHANGS-HST 和 LEGUS 之间系统差异的检查。颜色-颜色图中数据点的分布被用作“品质因数”，以进一步测试不同方法的相对性能。研究人员检查了使用不同聚类分类方法对科学结果（例如质量和年龄函数的确定）的影响，发现影响很小。
When completed, the PHANGS–HSTproject will provide a census of roughly 50 000 compact star clusters and associations, as well as human morphological classifications for roughly 20 000 of those objects. These large numbers motivated the development of a more objective and repeatable method to help perform source classifications. In this paper, we consider the results for five PHANGS–HSTgalaxies (NGC 628, NGC 1433, NGC 1566, NGC 3351, NGC 3627) using classifications from two convolutional neural network architectures (RESNET and VGG) trained using deep transfer learning techniques. The results are compared to classifications performed by humans. The primary result is that the neural network classifications are comparable in quality to the human classifications with typical agreement around 70 to 80 per cent for Class 1 clusters (symmetric, centrally concentrated) and 40 to 70 per cent for Class 2 clusters (asymmetric, centrally concentrated). If Class 1 and 2 are considered together the agreement is 82 ± 3 per cent. Dependencies on magnitudes, crowding, and background surface brightness are examined. A detailed description of the criteria and methodology used for the human classifications is included along with an examination of systematic differences between PHANGS–HSTand LEGUS. The distribution of data points in a colour–colour diagram is used as a ‘figure of merit’ to further test the relative performances of the different methods. The effects on science results (e.g. determinations of mass and age functions) of using different cluster classification methods are examined and found to be minimal.