A CATALOG OF VISUAL-LIKE MORPHOLOGIES IN THE 5 CANDELS FIELDS USING DEEP LEARNING
A CATALOG OF VISUAL-LIKE MORPHOLOGIES IN THE 5 CANDELS FIELDS USING DEEP LEARNING
复制标题
DOI：
10.1088/0067-0049/221/1/8
复制
发表时间：
2015-11-01
影响因子：
8.7
通讯作者：
Mcintosh, D. H.
中科院分区：
文献类型：
--
作者：
Huertas-Company, M.;Gravet, R.;Mcintosh, D. H.
我们在5个CANDELS场（GOODS-N，GOODS-S，UDS，EGS，和COSMOS）中提供了一个类似于50.000个星系（H-f160 w < 24.5）的类视H带形态的星表。使用卷积神经网络（ConvNets）估计形态。样品的中值红移< z >与1.25相似。该算法在GOODS-S上进行训练，其中视觉分类是公开的，然后应用于其他4个领域。根据CANDELS的主要形态分类方案，我们的模型检索每个星系的概率有一个球体或磁盘，呈现不规则性，紧凑或点源，是不可分类的。ConvNets能够以零偏差和类似于10%散射的方式预测星系图像的投票分数。错误分类的比例小于1%。我们的分类方案代表了相对于基于浓度-不对称-平滑度的方法的重大改进，该方法在高z下达到20%-30%的污染限制。
We present a catalog of visual-like H-band morphologies of similar to 50.000 galaxies (H-f160w < 24.5) in the 5 CANDELS fields (GOODS-N, GOODS-S, UDS, EGS, and COSMOS). Morphologies are estimated using Convolutional Neural Networks (ConvNets). The median redshift of the sample is < z > similar to 1.25. The algorithm is trained on GOODS-S, for which visual classifications are publicly available, and then applied to the other 4 fields. Following the CANDELS main morphology classification scheme, our model retrieves for each galaxy the probabilities of having a spheroid or a disk, presenting an irregularity, being compact or a point source, and being unclassifiable. ConvNets are able to predict the fractions of votes given to a galaxy image with zero bias and similar to 10% scatter. The fraction of mis-classifications is less than 1%. Our classification scheme represents a major improvement with respect to Concentration-Asymmetry-Smoothness-based methods, which hit a 20%-30% contamination limit at high z.