AI 中文总结
研究针对大量恒星样本年龄估计难的问题，开发不确定性感知的NGBoost框架，通过修改损失函数、采用蒙特卡罗策略，结合盖亚XP数据构建含千万颗恒星的年龄目录，给出年龄及不确定性估计。
AI 中文摘要
恒星年龄是银河系考古学的一个基本量，但对大量恒星样本进行可靠的年龄估计仍然具有挑战性。在这项工作中，我们开发了一个不确定性感知的NGBoost框架，用于使用盖亚XP衍生的大气参数和化学丰度进行恒星年龄估计。与标准NGBoost模型不同，我们通过纳入训练年龄标签的不确定性来修改损失函数。我们还使用蒙特卡罗策略来量化输入特征不确定性对预测年龄的影响。所得模型提供年龄估计以及不确定性估计。将此框架应用于盖亚XP恒星，我们构建了一个包含15,175,107颗恒星的恒星年龄目录。
英文摘要
Stellar age is a fundamental quantity for Galactic archaeology, but reliable age estimation for large stellar samples remains challenging. In this work, we develop an uncertainty aware NGBoost framework for stellar age estimation using Gaia XP-derived atmospheric parameters and chemical abundances. Different from the standard NGBoost model, we modify the loss function by incorporating the uncertainties of the training age labels. We further use a Monte Carlo strategy to quantify the influence of input-feature uncertainties on the predicted ages. The resulting model provides age estimates together with uncertainty estimates. Applying this framework to Gaia XP stars, we construct a stellar age catalog containing 15,175,107 stars.
CommentsAccepted to appear in the proceedings of the 9th International Conference on Pattern Recognition and Artificial Intelligence (PRAI 2026). 6 pages, 4 figures; stellar age catalog available online
