基于可行域解析中心的非线性回归

Non-linear Regression Based on Analytical Center of Feasible Region

提出了基于可行域解析中心的非线性回归算法,它克服了支撑向量回归因可行域不对称或狭长时其泛化性能降低的不足。从理论上分析了该回归算法与最大似然参数估计之间的关系,给出了它的迭代步骤,最后通过sinc函数的逼近问题验证了此回归算法的有效性。

Non-linear regression algorithm was proposed based on the analytical center of feasible reglon, which overcomes the drawback the generalization performance of Support vector regression degrades when the feasible space is elongated or asymmetric; the proof was given that the algorithm proposed is equivaient to the maximum likelihood estimation when fitting error meets a class of probability density function; The experiments validate that the non-linear regression proposed outperforms support vector regression.