超参数

而拟牛顿算法，可在一个校验集上最小化一个经验误差估计来优化SVM的超参数，使超参数在分类任务中达到最优值，从而取得理想的分类结果。

For optimizing SVM hyper-parameters , this optimization scheme minimizes an empirical error estimate using a quasi-Newton optimization method on the validation set .

而且模型的超参数都可以通过最大化边缘似然函数直接最优化得到，不需要使用到计算复杂的交叉验证技术。

The hyper-parameters are obtained easily by maximizing the marginal likelihood without resorting to expensive cross-validation technique .

支持向量回归学习方法中基于稳定型GA的超参数选择

Hyper Parameters Selection for SVR Based on Steady-State Genetic Algorithm

文中提出了一种基于稳定型遗传算法的支持向量回归模型（SVR）超参数选择方法。

This paper presents a new method to select the hyper parameters of SVR based on steady-state Genetic Algorithm ( GA ) .

作为一种具有很好应用前景的学习机器，支持向量机（SVM）的超参数选择在实际应用中一直是一个十分重要的问题。

As one of the promising learning machines , parameters selection is one of the key issues in the Support Vector Machine ( SVM ), specifically the Support Vector Regression ( SVR ), learning method .

本文的主要研究工作归纳如下：1.针对LS-SVM进行非线性建模时存在的超参数选取困难问题，提出一种基于混沌蚁群（CAS）算法优化的LS-SVM建模方法。

A novel LS-SVM modeling method is presented based on chaotic ant swarm ( CAS ) algorithm with the hyper-parameter selection using LS-SVM to model the nonlinear systems .

针对超参数选择困难的问题，采用多级网格搜索和遗传算法的方法对超参数的取值进行优化。

Due to the difficulty in parameter selection , the parameter was optimized using multi-grid searching method and genetic algorithm method .

我们也证明了序列稀疏贪婪优化的全局收敛性并讨论了如何利用它来优化多个超参数。

This paper also demonstrates the global convergence of SSGO and shows how to use it to perform model selection efficiently .

利用梯度算法，在超参数空间最小化支持向量机风险上界可以进行超参数的选取。

For choosing kernel hyper-parameters , the gradient method minimizes the risk under bound of support vector machines in hyper-parameters space .

在每个组内，可以进行信息共享并实施分层贝叶斯推理得出共享的超参数。

In each group , the Bayesian inference is used for information sharing and one common sparseness hyper-parameters is discovered in each group .

本文提出基于分布估计算法的高斯过程建模超参数优化算法，该算法能够克服局部搜索算法陷入局部最优的缺点。

Optimization of the hyper-parameters in GP surrogate modeling based on EDAs is proposed , which can solve the problem of local search algorithms .

同时，保持其他层次上的参数、超参数和模型参数固定不变，以简化贝叶斯前馈神经网络的参数估计。

Simultaneousty , the parameters , hyper parameters and model parameters at the other levels keep unchanged to simplify the parameters estimation of Bayesian feed forward networks .

有限元计算中使用的是8节点40自由度超参数壳体单元，计算结果和试验结果获得了令人满意的一致。

Superparametric shell element with eight nodes forty degrees of freedom is used in the calculation of finite element , the calculation results coincide with test results satisfactorily .

与此同时，把使用超参数温度壳元计算结果和相对自由度温度壳元计算结果相比较，表明了超参数温度壳元和相对自由度温度壳元有良好的一致性。

According to relative degree-of-freedom ( DOF ) theory , this dissertation also studies relative DOF thermal shell element , and compares the advantages of this element with others by numerical examples .

由于相关向量机的提出时间短，很多方面有待进一步的改进，尤其是算法中涉及到的矩阵求逆以及超参数的迭代算法，这大大的限制了相关向量机的应用范围。

Relevance vector machine was proposed recently , which needs to be further improved in many ways , particularly , Matrix inversion and iterative Solution of hyper parameters in the algorithm . These limit the scope of application of the relevant vector machine .

介绍了一类在固定人口条件下的标准流行病的随机模型，讨论了该模型的解及其性质，进一步对其参数进行了贝叶斯估计，对其先验分布的超参数也给出了确定的方法。

This paper analysed a kind of model of the general stochastic epidemic in regular population , discussed its solution and the solution 's property , and further carried out Bayesian estimation for its parameter , also gave a method of determining the super parameter of the prior distribution .

氧缺位对纳米SnO2喇曼谱及超精细参数的影响

Effects Produced by Oxygen Deficiencies on Raman Spectra and Hyperfine Parameters of Nano SnO 2

还提出了基于稳健回归算法的直流分量辨识算法，以及基于Hilbert变换和非线性变量优化（NLO）的基波分量辨识算法，实现了同步电机瞬态和超瞬态参数的精确辨识。

In addition , this article presents a technique using robust regression algorithm to analysis the DC current component , as well as a new algorithm based on Hilbert and NLO for the analysis of fundamental current component , consequently , realizes accurate identification of transient and subtransient parameters .

亚临界与超临界参数锅炉水动力特性的分析

Analysis of Water Dynamic Properties for Subcritical and Supercritical Pressure Boiler

超高压参数胀管锅炉酸洗问题的探讨

Probe on Acid Cleaning of Super High Pressure Boiler with Expanded Tubes

超临界参数机组控制系统的特点及其控制策略

Features and Strategies for Control Systems of Supercritical Pressure Sets

论发展超临界参数锅炉

On the Development of Boilers of Supercritical Parameters

超临界参数机组节能机理及应用

Energy Saving Mechanism and Application of Super-critical Units

超临界参数机组负荷生产过程是一个非线性、参数时变、多变量耦合的被控对象，在保证锅炉和汽轮机稳定运行的同时，还要考虑它们之间动态特性差异，实现锅炉、汽轮机协调控制。

The supercritical power unit load process is nonlinear , time-variation and multivariable coupling .

超临界参数燃煤机组的可用率剖析

Availability Analysis of Supercritical Coal-Fired Units

探讨了超临界参数锅炉的水动力特性及传热特性与工质热物理特性、水冷壁型式、运行方式的相关性。

It also discussessubstance thermal physics property water wall model and operation pattern of the supercritical pressure boiler .

分析了超临界参数单元机组的动态特性及各环节机理特性，概括了机组的建模研究现状。

The dynamics of supercritical power unit and mechanism of its components are analyzed and its modeling research status is presented .

以上结果表明，借助于此类电光晶体光瞳滤波器，不仅可以实现多功能光束整形及其实时操控，还可以避免传统光瞳滤波器一旦制成就不能再改变其光学超分辨参数的缺陷。

The results show that this kind of electro-optic pupil filter not only can realize multiple beam shaping and real-time manipulating , but also can overcome the limitation which the superresolution properties can not be adjusted with traditional pupil filter .

介绍了超临界参数机组负荷生产过程的主要输入输出变量及控制目标，推导了多变量广义预测控制与优化算法的基本形式及其约束处理。

The main manipulated variables and controlled outputs of supercritical power unit load process are introduced as well as control objects . A detailed analysis of multivariable generalized predictive control and optimization strategy and its constraints dealing is given . 3 .

超临界参数火电机组由于运行参数的提高，相比于同容量等级的亚临界机组热效率更高，在运行经济性和环保方面也有着巨大优势。

The thermal efficiency of supercritical thermal power units is higher than sub-critical units of the same capacity level , because of the improvement of the operating parameters . And also , supercritical units have great advantage in operating economy and environmental protection .