AI-assisted superresolution cosmological simulations
AI-assisted superresolution cosmological simulations
复制标题
DOI：
10.1073/pnas.2022038118
复制
发表时间：
2020-10
期刊：
影响因子：
--
通讯作者：
Yin Li;Y. Ni;R. Croft;T. Di Matteo;Simeon Bird;Yu Feng
中科院分区：
文献类型：
--
作者：
Yin Li;Y. Ni;R. Croft;T. Di Matteo;Simeon Bird;Yu Feng
宇宙学模拟对于理解我们的宇宙是不可或缺的，从宇宙网的创建到星系及其中心黑洞的形成。这种巨大的动态范围导致了巨大的计算成本，要求牺牲分辨率或尺寸，通常两者兼而有之。我们构建了一个深度神经网络来增强低分辨率暗物质模拟，生成超分辨率实现，这些实现在统计特性上与真实的高分辨率对应物非常一致，并且速度快了几个数量级。它很容易适用于更大的体积，并推广到训练数据中不存在的罕见对象。我们的研究表明，深度学习和宇宙学模拟可以成为一种强大的组合，可以在整个动态范围内模拟我们宇宙的结构形成。星系形成的宇宙学模拟受到有限计算资源的限制。我们借鉴人工智能（AI;特别是深度学习）的持续快速发展来解决这个问题。神经网络已经被开发来从高分辨率（HR）图像数据中学习，然后对不同的低分辨率（LR）图像进行精确的超分辨率（SR）版本。我们将这种技术应用于LR宇宙学N体模拟，生成SR版本。具体来说，我们能够通过生成512倍的粒子并预测它们从初始位置的位移来提高模拟分辨率。因此，我们的结果可以被视为模拟实现本身，而不是投影，例如，它们的密度场此外，生成过程是随机的，使我们能够采样的小尺度模式的大尺度环境条件。我们的模型仅从16对小体积LR-HR模拟中学习，然后能够生成SR模拟，成功地将HR物质功率谱复制到16 h− 1 Mpc的百分比水平，并将HR晕质量函数复制到10%以内，直到1011 M。我们成功地将模型部署在比训练模拟箱大1,000倍的箱子中，这表明可以快速生成高分辨率的模拟调查。我们的结论是，人工智能的帮助有可能彻底改变大宇宙学体积中小尺度星系形成物理学的建模。
Significance Cosmological simulations are indispensable for understanding our Universe, from the creation of the cosmic web to the formation of galaxies and their central black holes. This vast dynamic range incurs large computational costs, demanding sacrifice of either resolution or size and often both. We build a deep neural network to enhance low-resolution dark-matter simulations, generating superresolution realizations that agree remarkably well with authentic high-resolution counterparts on their statistical properties and are orders-of-magnitude faster. It readily applies to larger volumes and generalizes to rare objects not present in the training data. Our study shows that deep learning and cosmological simulations can be a powerful combination to model the structure formation of our Universe over its full dynamic range. Cosmological simulations of galaxy formation are limited by finite computational resources. We draw from the ongoing rapid advances in artificial intelligence (AI; specifically deep learning) to address this problem. Neural networks have been developed to learn from high-resolution (HR) image data and then make accurate superresolution (SR) versions of different low-resolution (LR) images. We apply such techniques to LR cosmological N-body simulations, generating SR versions. Specifically, we are able to enhance the simulation resolution by generating 512 times more particles and predicting their displacements from the initial positions. Therefore, our results can be viewed as simulation realizations themselves, rather than projections, e.g., to their density fields. Furthermore, the generation process is stochastic, enabling us to sample the small-scale modes conditioning on the large-scale environment. Our model learns from only 16 pairs of small-volume LR-HR simulations and is then able to generate SR simulations that successfully reproduce the HR matter power spectrum to percent level up to 16 h−1Mpc and the HR halo mass function to within 10% down to 1011 M⊙. We successfully deploy the model in a box 1,000 times larger than the training simulation box, showing that high-resolution mock surveys can be generated rapidly. We conclude that AI assistance has the potential to revolutionize modeling of small-scale galaxy-formation physics in large cosmological volumes.