【摘要】 针对拟合曲线时输入样本的随机性,提出了随机拟合的概念,并给出了相应的自适应算法.该算法能够自动学习拟合需要的统计特性,进而调整拟合多项式的系数.从理论上可以证明,采用该方法所得系数可以任意逼近加权最小二乘标准的系数.对于算法中区间的长度,进行了初步的分析及优化.仿真表明,该算法有较理想的逼近效果,优于一般的最小二乘法.更多还原

【Abstract】 The concept of random fit was brought forward and the adaptive algorithm was given. In this algorithm, the statistical properties of the input sequences are learned automatically, and coefficients of the polynomial can be updated correspondingly. It can be proved theoretically that the obtained coefficients can arbitrarily approach the optimal coefficients under the weighted least square criterion. Moreover, some analysis and optimization of the interval length were given.As demonstrated in simulation,this algorithm performs much better than the conventional least square method.更多还原