线性回归模型的误差项不服从正态分布或存在多个离群点时，可以将残差秩次的某些函数作为权重引入估计模型来减少离群点的不良影响。
When multiple outliers occur in linear regression model or the distribution of residuals is not normal, we can use residuals rank as weight function to get some resist estimator.
BP神经网络通过调节连接权重可以实现以任意精度逼近非线性函数，利用这一特点可对非线性函数关系进行拟合。
BP neural network can implement approximating nonlinear functions by arbitrary accuracy through regulating variable weight connection. This character can fit the nonlinear functions relations.
展望理论运用S状的价值函数和非线性的权重函数解释和预测框架效应。
An S-shaped value function and a nonlinear weighing function were employed by the prospect theory to explain and predict the framing effect.
应用推荐