Scholarlink

This paper introduces an astronomical spectral image dataset for benchmarking machine learning algorithms, named AstroSpectra-MNIST. It includes two versions: AstroSpectra-MNIST-v1 (stars, galaxies, quasars) and AstroSpectra-MNIST-v2 (F-type, G-type, K-type stars). Through a series of processes including data preprocessing and normalization, the astronomical spectral data from LAMOST are converted into lightweight grayscale images in the format of 28*28 pixels. We provide a lightweight, easily storable, and processable dataset. This paper demonstrates the benchmarking results of the dataset across various machine learning and deep learning models, verifying its effectiveness and challenges in astronomical spectral classification tasks. Future work will focus on data expansion and cross-domain applications.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

AstroSpectra-MNIST: An Astronomical Spectral Dataset for Benchmarking Machine Learning Algorithms

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Qiankun Wu,
  • Yungao Shi,
  • Ke Wang,
  • Ping Guo

摘要

This paper introduces an astronomical spectral image dataset for benchmarking machine learning algorithms, named AstroSpectra-MNIST. It includes two versions: AstroSpectra-MNIST-v1 (stars, galaxies, quasars) and AstroSpectra-MNIST-v2 (F-type, G-type, K-type stars). Through a series of processes including data preprocessing and normalization, the astronomical spectral data from LAMOST are converted into lightweight grayscale images in the format of 28*28 pixels. We provide a lightweight, easily storable, and processable dataset. This paper demonstrates the benchmarking results of the dataset across various machine learning and deep learning models, verifying its effectiveness and challenges in astronomical spectral classification tasks. Future work will focus on data expansion and cross-domain applications.