大学物理 ›› 2026, Vol. 45 ›› Issue (2): 98-.

• AI赋能 智教新探 • 上一篇    下一篇

基于人工智能符号回归的物理公式高效探索——以巴耳末公式为例

吕铁羽,吴顺情   

  1. 厦门大学 物理科学与技术学院,福建 厦门361005
  • 收稿日期:2025-03-09 修回日期:2025-05-23 出版日期:2026-05-15 发布日期:2026-05-21
  • 作者简介:吕铁羽(1975—),男,浙江缙云县,厦门大学物理科学与技术学院副教授,博士,主要从事大学物理教学和计算凝聚态物理研究工作
  • 基金资助:
    厦门大学2024年校级教学改革研究项目、厦门大学首批AI赋能课程教学改革创新校级项目资助

  • Received:2025-03-09 Revised:2025-05-23 Online:2026-05-15 Published:2026-05-21

摘要: 人工智能赋能的科学发现在重塑现代科学研究范式中展现突出潜力,将AI工具深度融入物理教学对于培养创新研究能力具重要意义.本文以氢原子巴耳末公式的发现过程为教学案例,采用符号回归技术实现物理规律在智能时代的重探索.通过构建巴耳末原始实验数据集,结合Feyn符号回归工具,系统展示从有限光谱数据重建特征公式的全过程.研究凸显“数据量质双驱动”原则,揭示人机协同模式下“数学发现”与“物理解释”的互补关系——AI虽可加速数学规律归纳,但对公式背后的物理解读仍需依托科学家的智慧.本案例实证揭示了人工智能工具加速数学规律归纳的有效性,同时更凸显物理思维引导模型迭代的关键支撑作用,为智能技术赋能物理教育创新提供一种有益思路.


关键词: 人工智能, 巴耳末公式, 符号回归

Abstract: Science discovery empowered by artificial intelligence (AI) demonstrates remarkable potential in reshaping the paradigm of modern scientific research. Deeply integrating AI tools into physics teaching is of great significance for cultivating innovative research capabilities. Taking the discovery process of the Balmer formula for the hydrogen atom as a teaching case, this paper employs symbolic regression technology to re-explore physical laws in the era of intelligence. By constructing the original experimental dataset of the Balmer series and combining it with the Feyn symbolic regression tool, the entire process of reconstructing the characteristic formula from limited spectral data is systematically demonstrated. The research highlights the principle of “dual-driven by the quantity and quality of data” and reveals the complementary relationship between “mathematical discovery” and “physical interpretation” under the human-machine collaboration model. While AI can accelerate the induction of mathematical laws, the physical interpretation behind formulas still relies on the wisdom of scientists. This case empirically reveals the effectiveness of AI tools in accelerating the induction of mathematical laws. At the same time, it more prominently demonstrates the key supporting role of physical thinking in guiding the iteration of the model, providing a useful idea for the innovation of physics education empowered by intelligent technology.

Key words:  artificial intelligence, Balmer formula, symbolic regression