arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2304.08192astro-ph.CO
收藏

宇宙值得 $64^3$ 个像素:用于宇宙学的卷积神经网络与 Vision Transformer

The Universe is worth $64^3$ pixels: Convolution Neural Network and Vision Transformers for Cosmology

Se Yeon Hwang, Cristiano G. Sabiu, Inkyu Park, Sungwook E. Hong

AI总结:

本文利用 $64^3$ 体素的暗物质晕三维光锥数据,比较 CNN、ViT 与 2pcf 的宇宙学参数估计能力,并通过融合方法与 Grad-CAM 解释提升精度和可理解性。

摘要:

我们提出了一种新方法,用于根据红移空间中暗物质晕的三维光锥数据估计宇宙学参数 $Ω_m$、$σ_8$、$w_0$ 以及一个派生参数 $S_8$;这些数据覆盖 $40^\circ\times40^\circ$ 的天区和 $0.3<z<0.8$ 的红移范围,并被分箱为 $64^3$ 个体素。我们使用两种深度学习算法——Convolutional Neural Network(CNN,卷积神经网络)和 Vision Transformer(ViT,视觉 Transformer)——并将它们的性能与标准的两点相关函数(2pcf)进行比较。结果表明,CNN 取得了最佳性能,而 ViT 在预测宇宙学参数方面也展现出显著潜力。通过结合 Vision Transformer、Convolutional Neural Network 和 2pcf 的输出,与单独使用 2pcf 相比,我们实现了误差的大幅降低。为了更好地理解机器学习算法的内部工作机制,我们采用 Grad-CAM 方法研究 CNN 和 ViT 激活图中关键信息的来源。研究结果表明,根据所预测参数的不同,这些算法会关注密度场和红移的不同部分。这项概念验证工作为将深度学习方法纳入从大尺度结构估计宇宙学参数铺平了道路,有望带来更紧的约束并增进对宇宙的理解。

英文摘要:

We present a novel approach for estimating cosmological parameters, $Ω_m$, $σ_8$, $w_0$, and one derived parameter, $S_8$, from 3D lightcone data of dark matter halos in redshift space covering a sky area of $40^\circ \times 40^\circ$ and redshift range of $0.3 < z < 0.8$, binned to $64^3$ voxels. Using two deep learning algorithms, Convolutional Neural Network (CNN) and Vision Transformer (ViT), we compare their performance with the standard two-point correlation (2pcf) function. Our results indicate that CNN yields the best performance, while ViT also demonstrates significant potential in predicting cosmological parameters. By combining the outcomes of Vision Transformer, Convolution Neural Network, and 2pcf, we achieved a substantial reduction in error compared to the 2pcf alone. To better understand the inner workings of the machine learning algorithms, we employed the Grad-CAM method to investigate the sources of essential information in activation maps of the CNN and ViT. Our findings suggest that the algorithms focus on different parts of the density field and redshift depending on which parameter they are predicting. This proof-of-concept work paves the way for incorporating deep learning methods to estimate cosmological parameters from large-scale structures, potentially leading to tighter constraints and improved understanding of the Universe.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑