宇宙暗物质模拟仿真数据集

宇宙暗物质模拟仿真数据集该数据集为基于HPC-AI框架移植的示范案例“HPC-AI宇宙学应用”的训练数据集，由HPC-AI宇宙学应用执行N-Body模拟程序产生宇宙暗物质分布数据，用于训练深度学习模型，预测宇宙物理参数。为了保证数据的客观准确性,由北京航空航天大学软件测评实验室于2024年6月进行第三方测试时产生，测试之前对测试条件、测试环境等进行了严格的检查和测量，确保了测试的有效性和数据的准确性。

宇宙暗物质模拟仿真数据集在HPC-AI协同开发和运行框架上使用MUSIC和pycola软件包生成暗物质的n-body模拟。MUSIC是一个生成宇宙初始条件的应用，MUSIC在模拟运行前随机初始化三个描述宇宙状态的参数：暗物质比率, 原始功率谱指数，物质超密度方差。pycola是一个多线程的Python/Cython n-body应用，可以根据MUSIC生成的初始数据来生成宇宙暗物质的位置分布。从pycola模拟中收集原始数据后，进行数据预处理，将原始数据转换为可以直接输入到3D CNN模型的训练数据。

Cosmic Dark Matter Simulation Dataset
This dataset is the training dataset for the demo case "HPC-AI Cosmology Application" ported based on the HPC-AI framework. It is generated by the N-Body simulation program of the HPC-AI Cosmology Application, which produces cosmic dark matter distribution data for training deep learning models to predict cosmic physical parameters. To ensure the objectivity and accuracy of the data, it was generated during the third-party testing conducted by the Software Testing Laboratory of Beihang University in June 2024. Prior to the test, strict inspections and measurements of test conditions and environments were carried out to guarantee the validity of the test and the accuracy of the data.

The Cosmic Dark Matter Simulation Dataset uses the MUSIC and pycola software packages on the HPC-AI collaborative development and operation framework to generate dark matter N-body simulations. MUSIC is an application for generating cosmic initial conditions. Before the simulation runs, MUSIC randomly initializes three parameters describing the cosmic state: dark matter fraction, primordial power spectrum index, and matter overdensity variance. pycola is a multi-threaded Python/Cython N-body application that can generate the positional distribution of cosmic dark matter based on the initial data generated by MUSIC. After collecting raw data from pycola simulations, data preprocessing is performed to convert the raw data into training data that can be directly input into 3D CNN models.

搜集汇总

背景与挑战

背景概述

该数据集基于HPC-AI框架，通过MUSIC和pycola软件包生成宇宙暗物质的n-body模拟数据，用于训练深度学习模型以预测宇宙物理参数。数据经过严格第三方测试和预处理，确保准确性，可直接输入3D CNN模型进行训练。