The most important criteria that used to check the calibrated model are root mean square error ( rms ) , the mean absolute error normalized rms error , and mass balance 模型参数使用试错法识别，识别过程中最重要的指标是均方差、平均绝对误差、标准均方差和水均衡。
Quantifying the performance of image denoising schemes by using the mean square error , the performance of the robust local threshold scheme is demonstrated and is compared with the universal threshold scheme 用均方差衡量去噪性能，实验表明用鲁棒局部阈值去噪法好于全局阈值去噪法。
The numerical experiments show that the proposed method can significantly raise snr , decrease mse and preserve the shape of signal better than the original wavelet shrinkage method 数值仿真表明，与小波阈值收缩方法相比，该方法在信噪比和均方差指标上都有明显的提高，并能很好地保持信号的形状。
The main work of this thesis includes two aspects : ( 1 ) modal identification based on wavelet transformation a equation of hard threshold wavelet de - noise is reasoned out from mse function 本文的主要工作分为两个方面： ( 1 )基于小波分析的模态识别根据最小均方差( mse )函数，本文推导了一个硬阈值小波消噪方程式。
To quantificationally analyze point spread function ( psf ) models for full digital autofocus method , an evaluation method of mean square error between original image and processed image was proposed for comparison performance of algorithms 摘要为了对用于全数字对焦的点扩散函数模型进行定量分析，提出用原始图像与处理所得图像的均方差进行性能评价的方法。
In the reservoirs with big fluctuation of inflow , the difference between lsm and robustified estimation methods ( including huber estimation and igg estimation ) is large and reduction extent of robust mean square error is large , and vice versa 入库流量波动大的水库，抗差估计方法与非抗差估计方法估计的参数差别较大，抗差均方差减小的幅度大；反之亦然。
The study on the distribution feature of global land rainfall for march - may during 1948 - 2001 points out that the maximum precipitation occurs in the monsoon areas , and their standard deviation is bigger in monsoon areas than the others 分析了全球陆地3 - 5月降水量的分布特点。指出降水量最大区域主要在季风区，而且季风区降水均方差较大，降水在纬带分布上有着较大的连贯性。
In the end , null shift and scale factor of the newly developed mdrlg is tested . random walk of the gyroscope is 0 . 0163 ? , rms of drift is less than 0 . 09 ? h , relative precision of scale factor is better than 10 ppm 最后对新研制的机抖陀螺进行了漂移和比例因子测试，实验测得陀螺随机游动系数为0 . 0163 / h ~ ( 1 / 2 ) ，漂移均方差小于0 . 09 / h ，比例因子相对精度优于10ppm 。
Finally , most supervised learning neural networks train themselves through minimizing mean squared error . but when the neural network models trained in this way are used to do forecasting , the existence of outliers result in great imprecision 最后，大多数监督学习神经网络是通过最小化训练集的均方差来训练网络，而野值的存在导致这种训练的神经网络模型在预测时会产生极大的不精确性。
At the same time , we proposed that the mean square errors of the last 5 numbers between the measured results and predicting results were used as the criterion that could judge the optimum model while we compared and analyzed several models ; we also compiled some programs for calculating 在模型的比选中，本文提出用实测与预测的最后五个值之间的均方差作为判别最优模型的标准，并编制了部分计算程序进行计算。
