Investigation of the regularization parameter of subspace-based optimization method for reconstruction of uniaxial anisotropic objects
Investigation of the regularization parameter of subspace-based optimization method for reconstruction of uniaxial anisotropic objects
复制标题
DOI：
10.1109/icsp.2016.7878155
复制
发表时间：
2016-11
期刊：
影响因子：
--
通讯作者：
Yulang Liu;Zhiqin Zhao;Xiaozhang Zhu;Z. Nie;Q. Liu
中科院分区：
文献类型：
--
作者：
Yulang Liu;Zhiqin Zhao;Xiaozhang Zhu;Z. Nie;Q. Liu
研究了正则化参数L对基于子空间的优化方法重建单轴各向异性物体性能的影响。得到了确定正则化参数L的一些准则。如果L太小，则成本函数可能不会很好地收敛。如果L太大，则噪声将被放大并且恶化逆过程。研究发现，单轴各向异性物体重建的SOM正则化参数与各向同性物体重建的SOM正则化参数具有相似的性质。
This paper investigates influence of the regularization parameter L on the performance of subspace-based optimization method in reconstructing uniaxial anisotropic objects. Some criteria are obtained to determine the regularization parameter L. If L is too small, the cost function may not converge well. If L is too large, the noise will be amplified and deteriorate the inverse process. It is found that the regularization parameter of SOM for reconstruction of uniaxial anisotropic objects behaves similarly as that in the isotropic case.