共查询到17条相似文献,搜索用时 140 毫秒
1.
2.
二进神经网络逻辑关系判据及等价性规则提取 总被引:3,自引:0,他引:3
二进神经网络中提取知识主要体现为对输入输出逻辑关系的提取,而逻辑关系的表达方式分为蕴含性规则和等价性规则文中对比了蕴含性规则和等价性规则的差异;以KT方法为例,讨论了蕴含性规则在表达二进神经网络内在知识时,对某些具有明确逻辑意义的二进神经网络,并不是最清晰的表达方式.对这些逻辑关系,采用等价性规则可以简洁清晰地解决问题,所以对于二进神经网络神经元表达的逻辑关系建立可能的等价性规则提取方法是有意义的.CH判据是一种提取等价性规则的方法,但CH判据是充分性判据,对二进神经元的权系数有约束条件,因此不适用于任何学习算法的学习结果.为解决这些问题,文中研究了二进神经网络表达几类等价逻辑关系的充要性判据,并根据这些判据提出了提取等价性规则的WTA方法.在使用WTA方法时,必须预先对二进神经元进行必要的剪枝.文中证明了剪枝定理,并通过二个例子说明了用WTA方法进行规则提取的过程. 相似文献
3.
在布尔空间中,汉明球突表达了一类结构清晰的布尔函数, 由于其特殊的几何特性,存在线性可分与线性不可分两种空间结构. 剖析汉明球突的逻辑意义对二进神经网络的规则提取十分重要, 然而,从线性可分的汉明球突中提取具有清晰逻辑意义的规则, 以及如何判定非线性可分的汉明球突,并得到其逻辑意义,仍然是二进神经网络研究中尚未很好解决的问题. 为此,本文首先根据汉明球突在汉明图上的几何特性, 采用真节点加权高度排序的方法, 提出对于任意布尔函数是否为汉明球突的判定算法;然后, 在此基础上利用已知结构的逻辑意义, 将汉明球突分解为若干个已知结构的并集,从而得到汉明球突的逻辑意义; 最后,通过实例说明判定任意布尔函数是否为汉明球突的过程, 并相应得到汉明球突的逻辑表达. 相似文献
4.
5.
6.
二进神经网络中笛卡尔球的研究 总被引:2,自引:0,他引:2
根据两类线性可分结构笛卡尔积的概念,定义了布尔空间中笛卡尔球的概念,证明了笛卡尔球是一类线性可分结构系.此外,还对以布尔空间中任意样本Xc为中心,与Xc之间Hamming距离为1的任意个样本与Xc组成的集合进行了研究,证明了这是一类笛卡尔球.为了对笛卡尔球进行规则提取,文中还分析了笛卡尔球的逻辑意义,建立了二进神经网络中判别笛卡尔球的一般方法,描述了这种判别方法的具体步骤,并通过一个实例说明了在二进神经网络中判别笛卡尔球的过程. 相似文献
7.
二进神经网络的汉明图学习算法 总被引:1,自引:0,他引:1
二进神经网络的几何学习算法ETL必须分隔全部真顶点集合和伪顶点集合,且为一种穷举的算法。该文使用所定义的汉明图和子汉明图,给出了选择核心顶点的依据,组成和扩展子汉明图的方向和步骤,以及一个子汉明图可用一个稳层神经元表示的条件和权值、阈值的计算公式。所提出的二进神经网络汉明图学习算法可用于任意布尔函数;无需穷举并一定收敛,因而是快速的;对文献所举实例取得了较ETL算法结构更为优化的三层前向网络。 相似文献
8.
在前台电力数据逻辑结构规则提取过程中,由于生成的规则非最优化,导致规则的表达能力不强、规则识别较低。基于此,设计一种前台电力数据逻辑结构规则自动提取方法。采用神经网络模型清洗数据,选择所需数据集,将其转换为满足神经网络要求的数据,并对其作进一步转化。针对转化后的数据,利用蚁群算法不断优化,实现对前台电力数据逻辑结构规则的自动提取。实验结果表明,与三种传统的规则提取方法相比,提出的规则自动提取方法所用规则长度最短,所需规则数量最少,规则的覆盖度最大可达到100%。 相似文献
9.
10.
稳健二进神经网络的几何训练 总被引:1,自引:0,他引:1
在二进神经网络的规划划分训练方法基础上,针对稳键二进神经网络稳健神经元的特点,讨论了稳健分类超平面的几何构造方法,并提出了相应的训练算法,包括隐层神经元的几何训练和输出神经无的进化训练。实验表明,该算法对复杂Boole函数的稳健二进神经网络的实现是有效且可行的。 相似文献
11.
We study the complexity of processing a class of rules called simple binary rule sets. The data referenced by the rules are stored in secondary memory. A necessary and sufficient condition that a simple binary rule set can be processed in a single pass of a file containing the base relations is given. Because not all simple binary rule sets can be processed in a single pass, a necessary and sufficient condition that a simple binary rule set can be processed by a constant number of passes is also given 相似文献
12.
The upper bound of the minimal number of hidden neurons for the parity problem in binary neural networks 总被引:1,自引:0,他引:1
Binary neural networks (BNNs) have important value in many application areas.They adopt linearly separable structures,which are simple and easy to implement by hardware.For a BNN with single hidden layer,the problem of how to determine the upper bound of the number of hidden neurons has not been solved well and truly.This paper defines a special structure called most isolated samples (MIS) in the Boolean space.We prove that at least 2 n 1 hidden neurons are needed to express the MIS logical relationship in the Boolean space if the hidden neurons of a BNN and its output neuron form a structure of AND/OR logic.Then the paper points out that the n -bit parity problem is just equivalent to the MIS structure.Furthermore,by proposing a new concept of restraining neuron and using it in the hidden layer,we can reduce the number of hidden neurons to n .This result explains the important role of restraining neurons in some cases.Finally,on the basis of Hamming sphere and SP function,both the restraining neuron and the n -bit parity problem are given a clear logical meaning,and can be described by a series of logical expressions. 相似文献
13.
Hopfield网络中二元正交记忆模式的吸引域分析 总被引:3,自引:0,他引:3
在作为联想记忆的Hopfield网络中,二元正交记忆模式的分析对网络记忆容量的研究起着重要作用。文中提出了利用吸引指数的概念对各个二元正交记忆模式的吸引域进行估计的方法。理论分析和计算机仿真表明,当网络容量不超过0.33N时(比通常的0.15N要好),每个二元正交记忆模式的吸引域至少包含一个汉明球。 相似文献
14.
RNA序列的高维空间二进制编码有以下优点:除可以对RNA序列的碱基结构、功能基团、碱基互补、氢键强弱等性质进行编码之外,还可方便的进行数学与逻辑运算。该文研究了RNA序列高维空间数字编码的更一般、更深刻的运算法则:(1)进一步研究RNA序列高维空间的表观维数NV,数值维数NX以及差异维数Nd,具体刻给出了当Nd=0,1,2,2n或2n+1(n=0,1,2,…)时,RNA序列的首段碱基及其数值取值范围。(2)推导出RNA序列多点“突变”(单核苷酸多态性SPN)的运算法则,将以前的结果推广到一般情形,深刻探讨了RNA序列之汉明距离、汉明值的变化及其数值变化情况。(3)利用RNA序列的定值部Xi和定位部Wi及其计算公式,从新的角度导出RNA重复序列的编码法则和运算法则,进而统一了以前的结果。 相似文献
15.
16.
Constraint Simplification Rules (CSR) is a subset of the Constraint Handling Rules (CHR) language. CHR is a powerful special-purpose declarative programming language for writing constraint solvers. The CSR subset of CHR forms essentially a committed-choice language consisting of guarded rules with multiple heads that replace constraints by simpler ones until they are solved. This paper gives declarative and operational semantics as well as soundness and completeness results for CSR programs.We also introduce a notion of confluence for CSR programs. Confluence is an essential syntactical property of any constraint solver. It ensures that the solver will always compute the same result for a given set of constraints independent of which rules are applied. It also means that it does not matter for the result in which order the constraints arrive at the constraint solver.We give a decidable, sufficient and necessary syntactic condition for confluence of terminating CSR programs. Moreover, as shown in this paper, confluence of a program implies consistency of its logical meaning (under a mild restriction). 相似文献
17.
A structural implementation of a fuzzy inference system through connectionist network based on MLP with logical neurons connected through binary and numerical weights is considered. The resulting fuzzy neural network is trained using classical backpropagation to learn the rules of inference of a fuzzy system, by adjustment of the numerical weights. For controller design, training is carried out off line in a closed loop simulation. Rules for the fuzzy logic controller are extracted from the network by interpreting the consequence weights as measure of confidence of the underlying rule. The framework is used in a simulation study for estimation and control of a pulp batch digester. The controlled variable, the Kappa number, a measure of lignin content in the pulp, which is not measurable is estimated through temperature and liquor concentration using the fuzzy neural network. On the other hand a fuzzy neural network is trained to control the Kappa number and rules are extracted from the trained network to construct a fuzzy logic controller. 相似文献