四元数及八元数线性空间中的内积及性质

内积空间理论是重要的数学基础,同时有非常丰富的应用背景.但通常的内积运算都是建立在实数域或复数域的线性空间上.当数域是四元数或八元数时,尚未有相应的内积空间理论.尝试在四元数及八元数线性空间中定义内积运算,并给出了若干性质.     

八元数上行列式定义及性质

八元数是一种非交换非结合代数，如何在八元数上建立行列式理论备受关注。用两种方法定义了八元数O上行列式，得到了一些性质及八元数三阶行列式计算机算法，还得到八元数二阶线性方程组的一般解法。     

新四元数系

与“正统”的Hamilton四元数不同，按作者的n元数运算统一规律，详细列举了新的四元数运算公式；如同对三元数的讨论方式，引进四元数的特征变换，论证了四元数特征与四元数的一致对应关系，从而得到四元数运算的另一等价形式即特征形式，据此可明了四元数与实数，复数以及三元数之间的密切联系，利用四维算术空间的特征轴和特征面，阐明了四元数运算的几何意义，利用引进的四元数的权值概念，建立了四元数的乘积定律，通过与Hamilton四元数运算的比较，确立了新四元数应有的地位。     

四元数矩阵的行列式的一种新定义及其应用

利用四元数矩阵的一种实表示,给出了四元数矩阵的行列式的一种定义及四元数矩阵的伴随矩阵的概念,讨论了四元数矩阵的行列式与伴随矩阵的性质,将四元数矩阵的这两个问题转换成实数矩阵的相应问题加以解决.     

三十二元数乘法表

给出三十二元数单元乘法表,并证明了能够表成32个平方数之和的两个数的乘积,也可以表成32个平方数之和.     

四元数矩阵特征值的估计定理

四元数矩阵的研究是矩阵理论和工程应用中的基本问题之一．本文研究了四元数体上矩阵的特征值的估计问题．给出了自共轭四元数矩阵特征值的最小最大值定理,得到了两个自共轭四元数矩阵的特征值之间的不等式．     

四元数矩阵两个不等式的改进

本文改进了文《关于四元数自共轭矩阵迹的几个不等式》中关于四元数自共轭矩阵迹的两个不等式.     

四元数向量和矩阵的实表示

在四元数和四元数向量、矩阵空间上引入3种不同的实数表示方式．将四元数之间及四元数向量和矩阵之间的运算，化为实数域上向量与矩阵之间的运算．得到的计算结果可准确转换成四元数和四元数向量和矩阵．可以在很大程度上克服四元数之间因乘积不可交换性而造成的计算困难．为计算机处理四元数数据提供可行的操作方案．     

双边四元数傅里叶变换的性质

重新证明了一个二维实函数的双边四元数傅里叶变换具有四元埃尔米特性，并利用此性质构造性地证明了一个二维实值函数的双边四元数傅里叶变换可以被分解成具有不同对称性的4个部分．     

四元数Z分布及其相关性质

在前人的基础上给出了四元数Z分布的定义,继而推导出了四元数Z分布的密度函数及其一些性质,并将得到的四元数Z分布的一些结果进行了推广．     

Conditional automaticity in retrieving multiplication facts: Evidence from Chinese

Adults automatically retrieve multiplication facts in a number matching task, in which participants judge whether the target number is one of the two cue digits presented previously. They are slower in rejecting the number which is the sum or the product of cues than a neutral number. The present study explored, with Chinese adult participants, whether such automaticity in retrieving multiplication facts can be affected by experiential and experimental factors such as learning experience and task set. We classified cue digits into two categories: small number first (the ascending order) and large number first (the descending order), according to whether the smaller or larger digit was presented on the left of the other. Experiment 1 found that product numbers were more difficult to reject than non-product numbers. Moreover, the order of cues played a role in modulating the magnitude of the interference effect, with ascending order conditions having larger interference than descending order conditions. Experiment 2 added a task-irrelevant addition or subtraction sign between the two cues and replicated the pattern of effects in Experiment 1, although the interference effects were significantly reduced. Experiment 3 instructed participants to judge whether the target number was the sum of the two cue digits. No interference effect was found for the target that was the product of cues. These findings suggest that the retrieval of multiplication facts, albeit automatic, can be influenced by learning experience (cue order effects in Experiments 1 & 2), the compatibility of the cuing context with the stored template information (reduced interference effects across Experiments 1 & 2), and task set (elimination of interference effects in Experiment 3). Retrieving multiplication facts is therefore conditionally automatic.     

Conditional automaticity in retrieving multiplication facts: Evidence from Chinese

Adults automatically retrieve multiplication facts in a number matching task, in which participants judge whether the target number is one of the two cue digits presented previously. They are slower in rejecting the number which is the sum or the product of cues than a neutral number. The present study explored, with Chinese adult participants, whether such automaticity in retrieving multiplication facts can be affected by experiential and experimental factors such as learning experience and task set. We classified cue digits into two categories: small number first (the ascending order) and large number first (the descending order), according to whether the smaller or larger digit was presented on the left of the other. Experiment 1 found that product numbers were more difficult to reject than non-product numbers. Moreover, the order of cues played a role in modulating the magnitude of the interference effect, with ascending order conditions having larger interference than descending order conditions. Experiment 2 added a task-irrelevant addition or subtraction sign between the two cues and replicated the pattern of effects in Experiment 1, although the interference effects were significantly reduced. Experiment 3 instructed participants to judge whether the target number was the sum of the two cue digits. No interference effect was found for the target that was the product of cues. These findings suggest that the retrieval of multiplication facts, albeit automatic, can be influenced by learning experience (cue order effects in Experiments 1 & 2), the compatibility of the cuing context with the stored template information (reduced interference effects across Experiments 1 & 2), and task set (elimination of interference effects in Experiment 3). Retrieving multiplication facts is therefore conditionally automatic.     

256比特以下整数乘法的快速实现

提出一种新的适用于256比特以下的整数乘法的软件实现方式,用软件实现大整数乘法时,一般采用所谓“纸笔运算”的方式,这种方式要求在内存中开辟一个区域来存放运算的中间结果,新的实现方式调整了乘法运算的步骤,充分地利用了寄存器组,几乎不需要用内存来存放中间结果,有效地减少了对内存访问的次数,从而提高了速度。     

矩阵乘法的并行计算及可扩展性分析

针对科学与工程计算中的大型矩阵乘积，与原有的串行算法相比照，利用向量外积法，构造了一种适用于多处理机系统的并行算法，给出了算法及实现过程，并进一步分析了算法的可扩展性．     

二进制域滑动窗口快速模乘算法

加速GF（2^m）上的模乘运算是提高GF（2^m）上ECC算法性能的关键。分析了窗口宽度ω的comb多项式乘法和NIST约简多项式算法,结合两种算法提出一种滑动窗口的快速模乘算法。该算法弥补了先乘后模方法在时间上和存储空间上的缺点,缩短了运算时间,仿真结果表明快速模算法的运算效率比先乘后模的窗口comb多项式乘法和NIST快速约简速度提高21%左右,预计算只需要ω-1个值。     

一种快速乘法算法--Karatsuba乘法算法

分析了一种快速乘法算法即Karatsuba算法的原理，比较了它对传统算法所具有的优越性，提出并论证了关于该算法的一些定理，指出了快速算法在现代计算机代数系统中的重要地位。     

基于Mastrovito乘法的字串行特征二域乘法器

针对目前常用的最低字优先字串行特征二域多项式基乘法器存在冗余计算的问题,提出了一种更加高效的最低字优先字串行乘法器。首先讨论了多项式模乘和Mastrovito乘法与最高位优先和最低位优先位串行乘法之间的关系,然后根据讨论发现的结果,将Mastrovito乘法器转变为字串行的形式,推导出新的最低字优先字串行乘法器。对综合所得的门级网表的比较显示:该乘法器的面积延时积比目前常用的最低字优先字串行乘法器小6.16%,比常用的最高字优先字串行乘法器小2.69%。     

Montgomery模幂运算的一种改进方案

在RSA算法中,大数模幂运算的核心是大数模乘运算。本文在传统的Montgomery算法的基础上,利用快速大整数平方运算,提出了Montgomery算法的一种改进方案,有效缩短了大数模幂运算的时间,从而提高了RSA算法的加解密速度。     

Fast multiplication on certain elliptic curves

The multiplication of points on elliptic curves is the most important operation in the implementation of elliptic curve cryptosystems. Based on Frobenius map, a fast multiplication on the curves defined by y" + xy = x + x~ + 1 over finite fields of characteristic 2 is given, and its optimality in the sense of using minimal numbers of additions of points is proved.     

基于存储的矩阵乘积优化算法

提出了一种基于存储的矩阵乘积优化算法.该算法转置矩阵,提高cache命中率,从而降低矩阵乘积时间. 实验结果表明此算法是行之有效的.