中文摘要：

连续域分布估计算法普遍采用高斯概率模型，假设变量服从高斯分布。该假设并不具有普遍意义。提出一个任意分布的连续多变量耦合分布估计算法，利用经验分布函数从样本估计分布，采样产生新的个体。描述经验分布函数和逆变换法采样，讨论用样本构造经验分布函数并采样的基本思想，给出一次采样算法及完整的分布估计算法，通过典型函数的仿真实验，说明方法的正确性和有效性。

英文摘要：

Estimation of distribution algorithms in continuous domains is based on such assumption that the variables subject to Gauss distribution.But it does not apply to everywhere.Estimation of distribution algorithms based on empirical distribution function is presented.The distribution is estimated with empirical distribution function directly.New individuals are sampled from the decided distribution for next generation.When the ideas of empirical distribution function and inverse transformation sampling are given,this paper describes the sampling algorithm and the whole estimation of distribution algorithm in order.The experimental results indicate the validity of the algorithm.