Skip to main content
QUICK REVIEW

[论文解读] HOLISMOKES -- XI. Evaluation of supervised neural networks for strong-lens searches in ground-based imaging surveys

R. Canameras, S. Schuldt|arXiv (Cornell University)|Jun 5, 2023
Pulsars and Gravitational Waves Research被引用 6
一句话总结

本研究评估了监督式深度神经网络在地面成像巡天中自动检测强引力透镜的能力,基于真实HSC PDR2数据(包含189个已确认的透镜和70,910个非透镜星系)。模型在零假阳性条件下实现了最高60%的真正例率,通过PSF感知训练和多波段数据实现了对图像质量近乎完美的不变性,为未来如Rubin LSST的巡天实现稳健、无需人工干预的透镜选择提供了可能。

ABSTRACT

While supervised neural networks have become state of the art for identifying the rare strong gravitational lenses from large imaging data sets, their selection remains significantly affected by the large number and diversity of nonlens contaminants. This work evaluates and compares systematically the performance of neural networks in order to move towards a rapid selection of galaxy-scale strong lenses with minimal human input in the era of deep, wide-scale surveys. We used multiband images from PDR2 of the HSC Wide survey to build test sets mimicking an actual classification experiment, with 189 strong lenses previously found over the HSC footprint and 70,910 nonlens galaxies in COSMOS. Multiple networks were trained on different sets of realistic strong-lens simulations and nonlens galaxies, with various architectures and data pre-processing. The overall performances strongly depend on the construction of the ground-truth training data and they typically, but not systematically, improve using our baseline residual network architecture. Improvements are found when applying random shifts to the image centroids and square root stretches to the pixel values, adding z band, or using random viewpoints of the original images, but not when adding difference images to subtract emission from the central galaxy. The most significant gain is obtained with committees of networks trained on different data sets, and showing a moderate overlap between populations of false positives. Nearly-perfect invariance to image quality can be achieved by training networks either with large number of bands, or jointly with the PSF and science frames. Overall, we show the possibility to reach a TPR0 as high as 60% for the test sets under consideration, which opens promising perspectives for pure selection of strong lenses without human input using the Rubin Observatory and other forthcoming ground-based surveys.

研究动机与目标

  • 开发并评估监督式神经网络,用于从地面成像巡天中自动、高精度地筛选星系尺度的强引力透镜。
  • 最小化透镜检测中的假阳性,尤其是在图像质量变化和观测效应(如PSF失配)的情况下。
  • 识别最优的网络架构、数据预处理和训练策略,以在保持零或极低假阳性率的同时最大化真正例率。
  • 评估多波段成像、PSF建模和数据增强对泛化能力和抗视宁度变化鲁棒性的影响。
  • 为未来的深空广域巡天(如Rubin LSST)实现完全自动化、高吞吐量的透镜选择流水线。

提出的方法

  • 在HSC PDR2巡天的真实强透镜模拟数据和非透镜星系上训练多种卷积神经网络,主要使用$gri$波段数据。
  • 采用多种架构(包括残差网络(ResNets)、类似AlexNet的模型和G-CNNs),以比较不同设计选择下的性能表现。
  • 应用数据增强技术,如随机图像质心偏移、平方根像素值变换和随机视角旋转,以提升鲁棒性。
  • 测试了加入$z$波段数据以及使用$g-\alpha i$差值图像($\alpha$经调优)以抑制中心星系辐射的效果。
  • 采用基于委员会的集成学习方法,对不同数据子集训练的网络进行组合,以减少假阳性重叠并提升泛化能力。
  • 在透镜模拟流程中集成真实的PSF模型,并将网络联合训练于科学帧和PSF剪裁图像,以实现对视宁度变化的不变性。
Figure 1: Mocks in our baseline ground truth data set. The first, second, and third rows contain mocks with $0.75\arcsec<\theta_{\rm E}<1.30\arcsec$ , $1.30\arcsec<\theta_{\rm E}<1.90\arcsec$ , and $1.90\arcsec<\theta_{\rm E}<2.50\arcsec$ , respectively. Cutouts have sizes of 10″ $\times$ 10″.
Figure 1: Mocks in our baseline ground truth data set. The first, second, and third rows contain mocks with $0.75\arcsec<\theta_{\rm E}<1.30\arcsec$ , $1.30\arcsec<\theta_{\rm E}<1.90\arcsec$ , and $1.90\arcsec<\theta_{\rm E}<2.50\arcsec$ , respectively. Cutouts have sizes of 10″ $\times$ 10″.

实验结果

研究问题

  • RQ1在真实强透镜数据上训练监督式神经网络时,零假阳性条件下的最大真正例率能达到多少?
  • RQ2在受控测试条件下,不同神经网络架构(如ResNet与类似AlexNet的模型)在强透镜检测中的性能表现如何比较?
  • RQ3数据预处理和增强技术在多大程度上能提升泛化能力和对图像质量变化的鲁棒性?
  • RQ4多波段训练(尤其是包含$z$波段)在减少非透镜形态污染方面的效果如何?
  • RQ5在不同数据子集上训练的集成模型能否显著降低假阳性重叠并提升检测召回率?

主要发现

  • 表现最佳的网络在零假阳性条件下实现了60%的真正例率(TPR 0),证明了纯自动化筛选的可行性。
  • 基于ResNet的架构始终优于类似AlexNet和G-CNN的模型,其TPR 0值达到约10%至40%。
  • 随机图像质心偏移和平方根像素值变换显著提升了性能,尤其在减少由PSF伪影引起的假阳性方面效果明显。
  • 加入$z$波段数据,并联合训练科学帧与PSF剪裁图像,显著提升了对视宁度变化的鲁棒性。
  • 添加$g-\alpha i$差值图像并未提升性能,且在抑制中心星系辐射方面无效。
  • 在不同数据集上训练、假阳性重叠适中的网络委员会模型实现了最高整体性能,凸显了集成多样性的重要价值。
Figure 2: Influence of the training data set for our baseline CNN (left, solid lines) and ResNet (right, dashed lines). We only vary the set of positive (top) and negative (bottom) examples. Networks trained on the baseline data sets (N1+L1) are plotted in dark blue. Two of the best networks from th
Figure 2: Influence of the training data set for our baseline CNN (left, solid lines) and ResNet (right, dashed lines). We only vary the set of positive (top) and negative (bottom) examples. Networks trained on the baseline data sets (N1+L1) are plotted in dark blue. Two of the best networks from th

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。