梯度法

The image pyramid is obtained by using the wavelet transform, and then we do the registering work at each level of the image pyramid. And the image grade method is applied at the coarsest level of the image pyramid, while Mutual information registration method can be performed at the rest of levels with the result at the pervious level serving as the initial condition.

通过小波变换构造图像金字塔，在每一层的平滑逼近分量上进行图像配准，并在金字塔最高层的近似图像上使用了梯度法进行图像配准，而在其余层上继续使用基于互信息的配准方法。

Based on the comparison of different algorithms of energy,we create a target function, and use grads method to realize localization.

本文在比较基于能量的各种算法的基础上提出了建立目标函数，用梯度法实现定位的算法。

The VSP-CDP imaging and Kirchhoof migration method were employed to get the fine profile of the upgoing wave.

采用最小平方共轭梯度法进行速度层析反演，得到井间反射波成像的速度模塑，利用VSP-CDP成像与克希正夫偏移相结合的方法对井间地震上行反射波进行精细成像。

The VSPCDP imaging and Kirchhoof migration method were employed to get the fine profile of the upgoing wave.

采用最小平方共轭梯度法进行速度层析反演，得到井间反射波成像的速度模型，利用VSPCDP成像与克希霍夫偏移相结合的方法对井间地震上行反射波进行精细成像。

The main role for the Line Up Competition Algorithm LCA is in global research , while Speediest Descent Algorithm SDA is in local search .

其中，列队竞争算法的主要作用是实现全局搜索，而梯度法的作用是实现局部寻优，这两种方法的结合，可以使全局搜索与局部搜索达到均衡和统一。

The main role for the Line-Up Competition Algorithm LCA is in global research, while Speediest Descent Algorithm is in local search. The computations show that the hybrid algorithm is superior to single one.

其中，列队竞争算法的主要作用是实现全局搜索，而梯度法的作用是实现局部寻优，这两种方法的结合，可以使全局搜索与局部搜索达到均衡和统一。

The inverse problem is solved using conjugate gradient method of minimization based on discrete ordinates method of radiative transfer equation.

该方法是在辐射传递方程离散坐标近似的基础上，用求目标函数极小值的共轭梯度法进行反演计算。

The real data verification shows that the wavelet, reflection factor and acoustic impedance given by PCGA are more accurate, quicker convergent and value stable than the existing direct sparse inversion method.

用实际数据检验算法，表明预条件共轭梯度法反演的子波、反射系数和波阻抗要比直接稀疏反演精度高、收敛快、数值稳定。

All the methods make the model more comply with the actual situation and avoid the onesidedness of single load operation condition.

根据分解协调原理，把这一复杂问题分解为最大负荷方式，一般负荷方式和最小负荷方式三个子模型，其投资费用在最大负荷方式下考虑，对于三个子模型用最大负荷损耗时间来协调统一，使得模型更符合实际情况，避免了只考虑单一负荷方式的片面性，具体求解方法采用非线性规划中的梯度法。

By defining a matrix function and introducing a free matrix U, the pole assignment constraint is relaxed, and the gradient σ2/U and σ2 σ2[V-1]/U are obtained. Thus robust design can be achieved by gradient method. Examples show that this method is very effective.

通过定义一矩阵函数并引入新的自由变量U，可放松极点配置的约束，并能系统的推导σ2/U及(σ2·σ2[V-1])/U，从而将鲁棒设计转化为无约束的梯度法寻优，实例说明，本文设计方法的效果很好。

Breath, muscle contraction of the buttocks; arch body, as far as possible to hold his head, right leg straight towards the ceiling (peg-leg knee in order to avoid muscle tension).

呼气，收缩臀部肌肉；拱起身体，尽量抬起头来，右腿伸直朝向天花板（膝微屈，以避免肌肉紧张）。

The cost of moving grain food products was unchanged from May, but year over year are up 8%.

粮食产品的运输费用与5月份相比没有变化，但却比去年同期高8％。