choosing a small regularization parameter or shortening the inversion range properly can help improve the inversion quality.
合理选择小正则化参数或者缩小反演范围能改善反演质量。
in the end, the genetic algorithms which has better precision and efficiency is adopted for finding the optimal regularization parameter based on the solution rule of regularization parameter.
最后根据正则化参数的确定原则,采用精度高和适应性更好的遗传算法确定最优正则化参数。
in order to solve the problem, we research into the regularization parameter and present a new method of adaptive selection of the regularization parameter (asrp).
针对这一问题,通过研究正则化参数的取值原理,提出了一种自适应选择正则化参数(asrp)的新方法。
so, a novel method for choosing the regularization parameter was presented. one numerical example was given with the comparison of generalized cross-validation (gcv) b-splines.
提出了一种选择调整参数的新方法,同时,给出了一个数学例子,并与广义交叉实验b—样条函数仿真比较验证。
numerical results show that 「near optimal」 parameter can be considered as an acceptable approximation of optimal regularization parameter with available priori information.
数值结果表明,在先验知识满足的条件下,近似最优参数法所找到的正则化参数是对最优正则化参数的较合理近似。
to obtain a good generalization performance, genetic algorithm is used to tune the regularization parameter and parameter of the kernel function when training the model.
为提高算法的推广性能,在模型训练过程中引入遗传算法自动选择惩罚因子和核函数宽度两个参数。
this paper utilizes l-curve method to determine the regularization parameter in the (above) two computation steps.
在两步计算中,均采用l曲线法来确定正则化参数α。
the regularization parameters are keys and l-curve method is used to fix on regularization parameter in regularization.
在正则化过程中,采用l曲线来确定正则参数,解决了正则参数的选取这一核心问题。
in this thesis, we choose truncated singular value decomposition to solve the resulting matrix equations, while the regularization parameter of tsvd is determined by the l-curve criterion.
鑒于此,必须采用正则化方法,本文中选用的是截断奇异值分解,其正则化参数用l-曲线準则来确定。
so the mathematical regularization methods were proposed to solve this problem, which made use of regularization parameter to achieve a balance between the noise and the true solution.
为解决这一问题,数学上提出了利用正则化参数在真值和噪声之间寻求平衡的正则化求解思想。