With the development of modern astronomical observation techniques and contact binary research, a large number of light curves of contact binaries have been published, and it has become a challenge to quickly derive the basic physical parameters of contact binaries from their light curves. This article presents a neural network (NN) based on the differential evolution intelligent optimization algorithm to infer the fundamental physical parameters of contact binaries from their light curve. Based on a large dataset of light curves and parameter data generated by Phoebe, a NN mapping model is established, while Differential Evolution (DE) and Markov Chain Monte Carlo (MCMC) algorithms are used to find reasonable parameter combinations, respectively. The experiments show that the parameter inversion speed of the DE algorithm is approximately 50% faster than that of the MCMC algorithm, while guaranteeing a parameter accuracy at least consistent with the those of MCMC algorithm.

中文翻译：

---

基于差分演化的接触双星参数快速反演方法

随着现代天文观测技术和接触双星研究的发展，大量的接触双星光变曲线被发表，从光变曲线快速推导接触双星的基本物理参数已成为一个挑战。本文提出了一种基于差分进化智能优化算法的神经网络（NN），用于从接触双星的光变曲线推断其基本物理参数。基于Phoebe生成的大型光变曲线数据集和参数数据，建立了神经网络映射模型，同时分别采用差分进化（DE）和马尔可夫链蒙特卡罗（MCMC）算法寻找合理的参数组合。实验表明，DE算法的参数反演速度比MCMC算法快约50%，同时保证参数精度至少与MCMC算法一致。