发表机构
École Polytechnique Fédérale de Lausanne (EPFL)（洛桑联邦理工学院）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
该研究针对星系团弥散射电辐射检测问题，提出集成散射变换与挤压-激励注意力的多分支分类架构，通过基准测试验证其在LoTSS-DR2/PSZ2数据集上的分类性能优于简单CNN，为SKA时代相关研究提供新方向。
AI 中文摘要
背景：星系团有时会在~100千秒差距至~1兆秒差距的尺度上产生同步辐射，其面亮度仅为图像噪声的数倍。这种弥散星系团射电辐射是探测磁场和星系团际介质动力学的灵敏探针，但要解析其潜在物理过程，需要涵盖广泛星系团质量、动力学状态和红移的统计大样本，同时对低面亮度辐射有足够灵敏度。目的：相对于基线分类器，探索两种提升星系团图像中弥散辐射检测的技术：散射变换（ST）和挤压-激励（SE）注意力。方法：将ST编码器集成到双分支分类器（DualSSN）和散射网络（ScatterNet）中；将SE注意力集成到DualSSN和双分支卷积神经网络（DualCSN）中。随后在10种图像预处理配置和3种裁剪策略下，将这些分类器与简单卷积神经网络（CNN）进行基准测试，性能评估基于LOFAR两米天空巡天第二次数据发布与普朗克Sunyaev-Zel'dovich源第二次星表重叠的小型标注数据集（LoTSS-DR2/PSZ2）。结果：结合SE与ST的多分支方法、将图像裁剪为固定数量望远镜波束、uv taper（平滑至更粗角分辨率）可提升分类性能，而堆叠图像的多个预处理版本则无提升。结论：基于散射变换的多分支架构结合波束归一化裁剪，是SKA时代弥散辐射分类的有前景方向。
英文摘要
Context. Galaxy clusters sometimes host synchrotron radiation on scales of approximately 100 kpc to approximately 1 Mpc, with a surface brightness only a few times the image noise. This diffuse cluster radio emission is a sensitive probe of magnetic fields and intracluster medium dynamics, but disentangling the underlying physical processes requires statistically large samples spanning a wide range of cluster masses, dynamical states, and redshifts, together with a sufficient sensitivity to low-surface-brightness emission. Aims. We explore two techniques for improving the detection of diffuse emission in galaxy cluster images, relative to a baseline classifier: the scattering transform (ST) and squeeze-excitation (SE) attention. Methods. We integrated an ST encoder into a dual-branch classifier (DualSSN) and a scattering network (ScatterNet). We incorporated SE attention into the DualSSN and dual-branch convolutional neural network (DualCSN). These classifiers were then benchmarked against a simple convolutional neural network (CNN), across ten image pre-processing configurations and three cropping strategies. Performance was evaluated on small labelled datasets from the second data release of the LOFAR Two-metre Sky Survey overlapping with the Second Planck catalogue of Sunyaev-Zel'dovich sources (LoTSS-DR2/PSZ2). Results. Alongside the multi-branch approach with SE and ST, cropping the image to a fixed number of telescope beams and uv tapering (smoothing to a coarser angular resolution) improve classification performance, while tacking multiple pre-processed versions of an image does not. Conclusions. Scattering-transform-based multi-branch architectures with beam-normalised cropping are a promising direction for diffuse emission classification in the SKA era.
Comments17 pages, 12 figures, Accepted in A&A
