一个基于傅里叶和小波的亚毫秒级模型，用于从NEOWISE单次曝光数据库中提取变源候选体
A Sub-Millisecond Fourier and Wavelet Based Model to Extract Variable Candidates from the NEOWISE Single-Exposure Database
AI总结：
本文提出基于小波分解与FEFT的深度学习模型VARnet，实现亚毫秒级天文光变曲线分析，在NEOWISE数据上高效提取变源，验证集F1达0.91且单源处理时间<53μs，为未来WISE完整变源巡天奠定基础。
摘要：
本文提出了VARnet，一个用于快速天文时间序列分析的强大信号处理模型。VARnet利用小波分解、一种通过有限嵌入傅里叶变换（FEFT）进行傅里叶特征提取的新方法以及深度学习来检测光变曲线中的微弱信号，利用现代GPU的优势实现亚毫秒级的单源运行时间。我们将VARnet应用于NEOWISE单次曝光数据库，该数据库包含10.5年间全天红外源的近2000亿个观测记录。本文设计了一个流水线，以便从NEOWISE数据中提取变源候选体，作为VARnet有效性以及对整个NEOWISE数据集进行未来变源巡天方法的可行性验证。我们实现了模型和模拟来合成独特的光变曲线以训练VARnet。在这种情况下，该模型在包含真实红外变源的验证集上，于4类分类方案中取得了0.91的F1分数。在配备22GB显存的GPU上，以每条光变曲线约2000个数据点计算，VARnet实现了小于53μs的单源处理时间。我们确认我们的VARnet对已知和此前未发现的变源均具有高灵敏度和精确度。这些方法为未来使用WISE进行完整的变源巡天展现了良好前景，并有效展示了VARnet模型架构的强大能力。
英文摘要：
This paper presents VARnet, a capable signal processing model for rapid astronomical timeseries analysis. VARnet leverages wavelet decomposition, a novel method of Fourier feature extraction via the Finite-Embedding Fourier Transform (FEFT), and deep learning to detect faint signals in light curves, utilizing the strengths of modern GPUs to achieve sub-millisecond single-source runtime. We apply VARnet to the NEOWISE Single-Exposure Database, which holds nearly 200 billion apparitions over 10.5 years of infrared sources on the entire sky. This paper devises a pipeline in order to extract variable candidates from the NEOWISE data, serving as a proof of concept for both the efficacy of VARnet and methods for an upcoming variability survey over the entirety of the NEOWISE dataset. We implement models and simulations to synthesize unique light curves to train VARnet. In this case, the model achieves an F1 score of $0.91$ over a 4-class classification scheme on a validation set of real variable sources present in the infrared. With $\sim2000$ points per light curve on a GPU with 22GB of VRAM, VARnet produces a per-source processing time of $<53μs$. We confirm that our VARnet is sensitive and precise to both known and previously undiscovered variable sources. These methods prove promising for a complete future survey of variability with WISE, and effectively showcase the power of the VARnet model architecture.
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。
