查询词典 regression

In order to decrease the error rate of speech recognition, speaker adaptation techniques are often used to adjust speaker-dependent acoustic models. MLLR (Maximum Likelihood Linear Regression) and MAP are two of the most popular techniques in recent years. MLLR uses the technique of regression trees. It calculates the transform matrix for each leaf node of the tree. This makes it possible to use fewer sentences to decrease the error rate of HMM-based speech recognition. However, while we examined the recognition result, we found that although the overall error rate decreased, but the error rate of certain confusable phones was higher. In order to solve this problem, we propose the use MPE (Minimum Phone Error Discriminative Training) to solve this problem.

在语音辨识技术的实际应用上，语者调适技术常被用做调整语者无关之声学模型，使其对於特定语者之辨识率有所提升，常见的语者调适技术有最大化相似度线性回归（Maximum Likelihood Linear Regression），其精神在於透过对语音辨识模型相近的mixture做群聚，再对各个群聚做调整，以期能达到利用少量语料，提升辨识率的目的，但其缺点为，在发音上相近的模型，其mixture原本就极为近似，若再分类为同一群聚做调整，容易因为使用者的发音习惯，而使模型偏向ㄓ或ㄗ，因此虽然整体的辨识率获得提升，但却造成混淆音的错误率上升。

The regression equations of B and Zn to the grein yield and the protein content in two fertilities have been set up with orthogonal polynomial regression method in this paper.

利用回归方法建立了两种肥力基础上产量和蛋白质含量硼、锌效应方程。

Upper molars showed low frequency of Carabelli cusps. Hypocone of upper second molars showed strong tendency of regression, but metacone had low regression tendency and a strong degree of compression.

上颚大臼齿部Carabelli cusp出现率极低，上颚第二大臼齿hypocone之退化倾向较强，metacone之退化情形较弱，而压扁程度较多。

On the basis of the theoretic analysis, this paper regards to the swatch of the listed corporate of our country in Shanghai Stock Exchange from 2002 to 2004, and sets up the model by linear regression and non-linear regression to analysis the listed company performance index sign and the financing structure, and carries through supposing examination, draws the conclusions:(1) The creditor's rights takes on the characteristic of a soft control and infirmness in the Chinese listed company governance, without bringing into play an obligatory function, the liability financing exists the best financing zone.

在理论分析的基础上，本文以2002―2004年度我国沪市上市公司为样本展开实证研究，采用线性回归和非线性回归、一元回归和多元回归等多种方法建立模型，进行上市公司绩效指标与融资结构的回归分析，并进行假设检验，得出如下结论：(1)债权在中国上市公司治理中表现出软约束和弱化的特征，没有发挥出应有的作用，债权融资存在最优融资区间。

I. An analysis of the prominently related variables with whether or not to receive cervical smear screen of tendency factors, enabling factors and needs factors by using the logistic regression model was performed.: 6 variables of tendency factors with predictive value as to whether a woman receive cervical smear screen or not include age, marital status, educational level, family structure, religion, and the knowledge of free smear examination offered by mandatory citizen health insurance; 4 variables with enabling factors with predictive value include average monthly income, district, whether receiving telephone or letter notices in the past year, and the degree of convenience of the examination location. One variable of needs factors with predictive value: is the suitability of the examination time notified. The accuracy rate of laddered logistic regression model built by this research was 74.1%.

经卡方检定后，将倾向、能用、需要因素中具有显著相关变项与是否接受抹片检查，放入逻辑斯回归预测模型中，本研究结果发现：倾向因素中的六个变项：年龄、婚姻状况、教育程度、家庭结构、宗教信仰、是否知道健保提供免费抹片检查对受检与否具有预测力；能用因素中的四个变项：平均月收入、地区别、过去一年内是否接到电话或信函通知、收到通知的受检地点是否方便对受检与否具有预测力；需要因素中的一个变项：收到通知的受检时间是否合适对受检与否具有预测力；本研究所建立的阶层逻辑斯回归模型的整体预测正确率为74.1﹪。

Using the Landsat Thematic Mapper data and the measured grass yield data,the monadic linear regression models and the non\|linear regression models were established,to express the relations between grassland biomass and the different vegetation indices.

通过陆地卫星TM遥感影像数据和同期野外调查数据，分析了植被指数与草地植被生物量之间的相关关系，建立了不同植被指数与草地生物量之间的一元线性回归模型和非线性回归模型。

Counting the "3 oxygen " and "3 nitrogen" data of the Dawangtan reservoir with the monadic linear regression method, by significance test of correlativity coefficient and regression coefficient, the author put forward the equation of correlativity significance tests.

运用一元线性回归对大王滩水库的&三氧&&三氮&资料进行统计，并经相关系数和回归系数显著性检验，得出相关性显著的方程，利用该方程可提高水质评价的准确性和监测效

The monodic regression equation of BCF and t_R and the binary regression equation T...

统计检验表明，两类估算方程在不同固定相上的平均相关系数分别为0·952和0·965，均呈极显著相关(P·01)，可用于估算取代苯类化合物在草鱼体内的BCF，估算结果不受固定相极性的影响。

In full-time model of the salary regression, this text with career planning regression analysis result be adjusted according to the multinominal logit model . The text find height taller female university student, possess taller full-time of the salary level, weight heavier female university student the full-time salary level then will be notable to reduce, when the female university student has self-confidence will be notable to promote full-time salary level. In addition match the of the shape's limit influence toward whole self-confidence and can also beg to be from a material for early "indirect" influence.

在全职薪资回归模型中，本文以生涯规划回归分析结果作为调整依据，即全职薪资的样本选择模型，本文发现身高较高的女大学生，其享有较高的全职薪资水准；体重较重的女大学生则全职薪资水准会显著降低；当女大学生有自信时，会显著提升全职薪资水准，此外搭配身材对於整体自信的边际影响，亦能求出身材对於初期全职薪资的「间接」影响。

Density Functional Theory method was used to optimize the geometries of 209 PCDE molecules, 209 PBDE molecules and 75 PCDD molecules at the B3LYP/6-31G* level. Using computed structural parameters as theoretical descriptors, the forward stepwise multiple regression technique was adopted to obtain QSPR models of environmental partition properties for these POPs by using GQSARF 2.0 and SPSS12.0 for windows programs. The obtained QSPR models are as follows:(1) QSPR models for predicting subcooled liquid vapor pressure, n-octanol/water partition coefficients lgK_(ow and subcooled liquid water solubilities -lgS_(w,l of PCDEs, of which correlative coefficients (R~2) are 0.988, 0.958 and 0.959 and the root-mean-square-error of estimation are 0.134, 0.116 and 0.327 respectively.(2) QSPR models of lgPL and n-octanol/air partition coefficients lgK_(oa for PBDEs, which both contain three structural parameters. The values of R2 for the two models are both 0.997 while the values of RMSEE are 0.073 and 0.062 respectively.(3) QSPR models of lgKow and–lgS_ for PCDEs, which both have one variable (mean molecular polarizability,α). The values of R~2 for the two models are 0.978 and 0.866 and the values of RMSEE are 0.300 and 0.270 respectively.(4) The molecular structures of 24 substituted naphthaline compounds were optimized using Hartree-Fock and DFT methods at four different levels and the same means was used to obtain four three-parameter (EHOMO, q~+ andα) QSPR models of lgK_. The model at the HF/6-311G** level is the best one of which R2 is 0.9662 and RMSEE is 0.380.(5) QSPR study for environmental partition properties of PCDEs was also performed using position of Cl substitution method in which simple parameters of substitution position were taken as descriptors. The multiple linear regression was performed with GQSARF 2.0 and SSPS 12.0 for windows programs to obtain QSPR models of lgP_L, lgK_ and–lgS_ for PCDEs of which R~2 are 0.991, 0.983 and 0.965 and RMSEE are 0.311, 0.100 and 0.300 respectively.

采用密度泛函理论方法在B3LYP/6-31G*水平上对PCDEs、PBDEs和PCDDs的分子结构进行了全优化计算，以计算得到的量子化学参数作为理论描述符，采用GQSARF 2.0和SPSS 12.0 for windows统计程序进行正向逐步回归分析，建立了这些POPs的环境分配性质的QSPR模型：(1) PCDEs的过冷液体蒸汽压、正辛醇/水分配系数lgK_(ow和水溶解度-lgS_(w，l的QSPR模型，这3个QSPR模型的相关系数(R2)分别为0.988、0.958和0.959，估计的均方根误差分别为0.134、0.116和0.327；(2) PBDEs的lgPL和正辛醇/空气分配系数lgK_(oa的QSPR模型，这两个模型都包含三个分子结构参数，其R~2都为0.997，RSMEE分别为0.073和0.062；(3) PCDDs类化合物的lgK_和-lgS_w的QSPR模型，两个模型都只含一个变量，其R~2分别为0.978和0.866，RSMEE分别为0.300和0.270；(4)采用Hartree-Fock和DFT方法，在4种不同水平上优化计算了24个取代萘系列化合物的分子结构，采用上述同样的方法分别建立了四种水平上的三变量lgK_模型，通过比较得到，在HF/6-311G**水平计算得到的模型最好，R为0.966，RSMEE为0.380；(5)同时，采用氯原子取代位置方法对PCDEs的环境分配性质进行QSPR研究，建立了PCDEs的lgPL、lgK_和-lgS_的QSPR模型，其R~2分别为0.991、0.983和0.965，RSMEE分别为0.110、0.100和0.300。

The teacher likes the honeymouthed little girl very much.

老师很喜欢这个嘴甜的小姑娘。

Mr. Notker Bien's interests are traveling, spending quality time with the family and long-distance-running.

诺特卡·柏恩先生热爱旅游，长跑，以及和家人一起共度美好时光。