一种新的支持向量机主动学习策略及其在文本分类中的应用

There are two well-known characteristics about text classification.One is that the dimension of the sample space is very high,while the number of examples available usually is very small.The other is that the example vectors are sparse.Meanwhile,we find existing support vector machines active learning approaches are subject to the influence of outliers.Based on these observations,this paper presents a new hybrid active learning approach.In this approach,to select the unlabelled example(s) to query,the learner takes into account both sparseness and high-di-mension characteristics of examples as well as its uncertainty about the examples‘‘ categorization.This way, the active learner needs less labeled examples,but still can get a good generalization performance more quickly than competing methods.Our empirical results indicate that this new approach is effective.     

Projections and preconditioning for inconsistent least-squares problems

In this paper we describe two iterative algorithms for the numerical solution of linear least-squares problems. They are based on a combination between an extension of the classical Kaczmarzs projections method (Popa [5]) and an approximate orthogonalization technique due to Kovarik. We prove that both new algorithms converge to any solution of an inconsistent and rank-defficient least-squares problem (with respect to the choice of the initial approxi-mation), the convergence being much faster than for the classical Kaczmarz - like methods. Some numerical experiments on a first kind integral equation are described.     

基于离散Hopfield神经网络的噪声数字识别

在Hebb学习规则的基础上,运用离散Hopficld神经网络的联想记忆能力,对含有噪声而产生畸变的0~9数字进行了识别。通过改进神经网络的记忆样本,即先对记忆样本做正交化处理,再对改进后的记忆样本进行学习,得到相应的权值矩阵,然后利用改进后的离散Hopfield神经网络根据待识别噪声数字的信息联想已记忆的数字。实验结果表明,改进后的神经网络对噪声数字有较好的识别效果,提高了记忆能力和识别的正确率。     

正交递归选择法及其在转炉炉龄研究中的应用

本文提出了一种新的变量筛选法—正交递归选择法,该法可以得到预报能力较强的模型,即PRESS(预报残差平方和)值较低的模型。转炉炉龄问题作为研究实例,并与逐步回归正向选择法及偏最小二乘回归法进行了比较。     

蒸发油汽吸收回收技术的研究(Ⅱ)--小型试验及其影响因素分析

油汽吸收分离操作及其效果受到许多因素的影响。利用拟水平正交设计方法安排试验并考察各影响因素。试验结果表明,吸收质、吸收剂、进气流量对综合吸收率 η Ｒ 及吸收剂流量 Ｑ Ｌ 的评分指标 ｙ 有显著的影响,吸收质、吸收剂对其影响较大。进口油汽浓度虽对 ｙ 的影响不显著,但对 分指标 η Ｒ和 Ｑ Ｌ 的影响分别为高度显著和较显著的。吸收剂 Ａ Ｂ Ｓ－ １ 、 Ａ Ｂ Ｓ－ ２ 的吸收效果相当且远好于煤油、轻柴油。最佳的试验条件（ 工艺参数） 为用吸收剂 Ａ Ｂ Ｓ－ １ 来吸收油汽体积分数为５０ ％ 、流量为１８０ ｍ Ｌ／ ｍｉｎ 的油气混合气。另外,从 Ａ Ｂ Ｓ－ １ 、 Ａ Ｂ Ｓ－ ２ 的对比吸收试验来看,除进一步说明吸收剂 Ａ Ｂ Ｓ－ １ 与 Ａ Ｂ Ｓ－ ２ 的吸收效果相当外,还认为试验中 Ａ Ｂ Ｓ－ １ 、 Ａ Ｂ Ｓ－ ２ 的用量低于国外同类回收装置的相应值。     

石墨缠绕垫片的结构与工艺参数的研究

本文研究了金属石墨缠绕垫片的结构和工艺参数对其性能的影响，其中引用了美国ＡＰＩ标准中的相关数据进行结构设计，从而提出了对工程实践具有指导意义的重要结论。     

Factoring ofx n−1 and orthogonalization over finite fields of characteristic 2

We present a new deterministic algorithm which gives a complete factorization ofx ⁿ – 1 over finite fields of characteristic 2. To do so we adapt a well-known procedure of orthogonalization in order to compute primitive idempotents over extensions instead of over GF(2).     

Modified Kovarik Algorithm For Approximate Orthogonalization Of Arbitrary Matrices

In a previous paper the author presented an extension of an iterative approximate orthogonalization algorithm, due to Z. Kovarik, for arbitrary rectangular matrices. In this algorithm, as Kovarik already observed in his paper, at each iteration an inversion of a symmetric and positive definite matrix is made. The dimension of this matrix equals the number of rows of the initial one, thus the inverse computation can be very expensive. In the present paper we describe an algorithm in which the above matrix inversion step is replaced by an arbitrary odd degree polynomial matrix expression. We prove that this new algorithm converges to the same matrix as the original Kovarik's method. Some numerical experiments described in the last section of the paper show us that, even for small degree polynomial expressions the convergence properties of the new algorithm are comparable with those of the original one.     

直接线性图嵌入算法及其在人脸识别中的应用

图嵌入算法使用无向有权图来描述数据集的流形结构,目前许多流形学习算法都可统一到这个框架下。线性图嵌入算法(LGE)在高维小样本应用中往往会遇到的奇异值问题,因此需把数据集预先投影到PCA子空间,往往会丢失了一些有用的信息。本文提出了一种直接的线性图嵌入算法(DLGE),可直接从原始数据集中提取特征。此外DLGE算法相对于基于迭代的正交化算法,在最小二乘意义下对截断的征向量进行正交化处理,计算简便有效。在多个人脸数据库库上的仿真结果表明,相对于传统算法,DLGE算法具有更强的人脸表征能力,更好的分类性能,且更加鲁棒。