A new neural network for solving linear programming problems andits application

Presents a new neural network which improves existing neural networks for solving general linear programming problems. The network, without setting parameter, uses only simple hardware in which no analog multipliers are required, and is proved to be completely stable to the exact solutions. Moreover, using this network the author can solve linear programming problems and its dual simultaneously, and cope with problems with nonunique solutions whose set is allowed to be unbounded.     

一种监督降维的OP-LASRC算法在行为识别中的应用*

针对大数据的人体行为识别时实时性差和识别率低的问题,提出了优化投影对线性近似稀疏表示分类(OP-LASRC)的监督降维算法。OP-LASRC将高维的行为数据优化投影到低维空间,与线性近似稀疏表示(LASCR)快速分类算法相结合应用大数据的人体行为识别。首先利用LASCR的残差计算规律设计OP-LASRC算法,实现监督降维;利用线性正交投影缩减高维数据的维度,投影时减小训练样本的本类重构残差及增大类间重构残差,从而保留训练样本的类别特征。然后,对降维后的行为数据,利用LASCR算法进行分类;用L2范数估算稀疏系数,选出前k个最大的稀疏系数对应的训练样本,缩减训练样本库后用L1范数最小化和残差最小化计算得到识别结果,从识别率、鲁棒性、执行时间三个方评价此方法,在KTH行为数据库上进行实验测试。实验表明：OP-LASRC监督降维后,LASRC在分类时不仅识别率高达96.5%,执行时间比同类算法短,而且保证了强鲁棒性,证明了OP-LASRC能完美匹配LASCR算法用于行为识别,这为大数据的行为识别提供了一种新的思路。     

基于几何矩的CAD模型形状匹配算法及应用

基于矩方法及其理论,提出了一个基于几何矩的形状匹配算法,用于识别CAD模型中具有相似形状特征的几何体.该算法采用一组满足平移、旋转、缩放不变性的几何矩不变量对三维几何体的形状特征进行描述,并根据形状特征向量的相似程度评估几何体之间的相似性.为提高几何矩计算的准确性和效率,对CAD模型的三角面片数据进行了预处理,并采用递...     

一种量子神经网络模型学习算法及应用

提出一种量子神经网络模型及学习算法. 首先基于生物神经元信息处理机制和量子计算原理构造出一种量子神经元, 该神经元由加权、聚合、活化、激励四部分组成. 然后由量子神经元构造出三层量子神经网络模型, 其输入和输出为实值向量, 权值和活性值为量子比特. 基于梯度下降法构造了该模型的超线性收敛学习算法. 通过模式识别和函数逼近两种仿真结果表明该模型及算法是有效的.     

Data analysis using a geometrical representation of predicate calculus

This paper proposes an approach to database analysis and design, utilizing predicate calculus in the “infospace,” i.e. in the space of the attributes. The difference B₁ − B₂ of two Boolean predicates in the canonical form, B₁ and B₂, is discussed, and a geometrical representation is given in an n-dimensional space. A tabular method is derived to help in query decomposition. The approach is the basis of the “infospatial derivative” (defined by the second author in [47,45,48]), and underlies the “projection-solid infospatial approach” to the decomposition of noncanonical-query retrieval into steps enjoying the consecutive-retrieval property [46,44]. Extension to expert database systems is investigated at the end, through a comparison with other trends: the universal relation, the relation-valued relational approach, and frame algebra.     

The geometrical learning of binary neural networks

In this paper, the learning algorithm called expand-and-truncate learning (ETL) is proposed to train multilayer binary neural networks (BNN) with guaranteed convergence for any binary-to-binary mapping. The most significant contribution of this paper is the development of a learning algorithm for three-layer BNN which guarantees the convergence, automatically determining a required number of neurons in the hidden layer. Furthermore, the learning speed of the proposed ETL algorithm is much faster than that of backpropagation learning algorithm in a binary field. Neurons in the proposed BNN employ a hard-limiter activation function, with only integer weights and integer thresholds. Therefore, this will greatly facilitate actual hardware implementation of the proposed BNN using currently available digital VLSI technology     

Design with part behaviors: behavior model, representation and applications

A part is primarily characterized by its ‘function(s)’. The function of the part is achieved through its ‘interactions’ with other parts in an assembly under a set of operating conditions. These ‘interactions’ constitute the ‘behavior’ of the part. The ‘part behavior’ is achieved through a set of spatial and design functional relationships between the interacting surfaces of the parts. The set of spatial and design functional relationships for a part constitutes the Part Function Model (PFM) of that part. In this paper, the nature and role of part ‘functions’ and ‘behaviors’ have been studied in the context of a product design system. The paper addresses the following issues: (i) relationship between function, behavior, and geometry of a part; (ii) importance of ‘part behavior’ over ‘part function’, and development of a ‘part behavior’ model; (iii) methodology for transforming the part behaviors into the PFM model of the part; (iv) product model framework for storing the PFM model information with an Object Oriented Programming (OOP) based CAD system (Concentra's Concept Modeller); (v) importance of the PFM model within the product development process; and (vi) application of the PFM model for generating various product specifications of the part. The prototype implementation of a Functional Design System for transforming part behaviors into different types of part specifications has also been presented in this paper.     

Symbolic representation of neural networks

Neural networks often surpass decision trees in predicting pattern classifications, but their predictions cannot be explained. This algorithm's symbolic representations make each prediction explicit and understandable. Our approach to understanding a neural network uses symbolic rules to represent the network decision process. The algorithm, NeuroRule, extracts these rules from a neural network. The network can be interpreted by the rules which, in general, preserve network accuracy and explain the prediction process. We based NeuroRule on a standard three layer feed forward network. NeuroRule consists of four phases. First, it builds a weight decay backpropagation network so that weights reflect the importance of the network's connections. Second, it prunes the network to remove irrelevant connections and units while maintaining the network's predictive accuracy. Third, it discretizes the hidden unit activation values by clustering. Finally, it extracts rules from the network with discretized hidden unit activation values     

Building a fuzzy logic-based McCulloch-Pitts Neuron recommendation model to uplift accuracy

The Journal of Supercomputing - Recommender system is one of the most popular technique used for information filtering. It helps in discovering hidden knowledge patterns from a large set of...     

EMD-BP神经网络预测模型及应用

时间序列分析是根据客观事物的连续性和规律性推测未来发展趋势的预测方法,分析时设法过滤除去不规则变动,突出反映趋势性和周期性变动。为了提高预测精度,构建了EMD-BP神经网络预测模型,利用Hilbert-Huang变换中的经验模态分解将时间序列分解为有限个本征模函数,重构后进行BP神经网络预测。通过对中国石化的股票资料进行实验仿真,表明该模型降低了被预测数据的非平稳性,其精度比直接用神经网络预测有较明显的提高。     

改进的非负稀疏编码神经网络模型及其应用

提出了一种改进的基于NIG（Normal Inverse Gaussian）密度和稳健主成分分析（PCA）的非负稀疏编码（NNSC）神经网络模型,该模型实质上实现了一个二阶段的学习过程。并利用这个模型成功地建模了视觉感知系统V1区的感受野。该NNSC模型具有很强的自适应于自然数据统计特性的能力。另外,利用类似小波收缩法去噪原理,该模型能够有效地去除图像中的高斯加性噪声,对自然图像编码的仿真实验也表明了该模型在生物学上的合理性和可行性。     

A parallel neural processor for real-time applications

Training and testing artificial neural networks can be challenging and time-consuming. Experiments with two real-time applications were performed to compare three approaches for implementing a multilayer perceptron neural network. In both applications, the special-purpose processor performed best     

A concurrent engineering-oriented design database representation model

A concurrent engineering-oriented design database representation model (CE-DDRM) is introduced in this research for supporting various life-cycle aspects in concurrent design. In this model, concepts and behaviors of different design database modeling components, including entities, properties, relationships, tasks, and specifications, are defined at meta-class level. Design database is modeled at two different levels, class level and instance level, representing generic design libraries and special design cases, respectively. A Web-based system architecture is proposed to model distributed design database and allow team-members for different product development life-cycle aspects at different locations to access the design database. This newly introduced approach provides the foundation for developing the next generation CAD systems with concurrent engineering functions.     

A new fuzzy neural network model for solving fuzzy linear programming problems and its applications

This paper is concentrated on two types of fuzzy linear programming problems. First type with fuzzy coefficients in the objective function and the second type with fuzzy right-hand side values and fuzzy variables. Considering fuzzy derivative and fuzzy differential equations, these kinds of problems are solved using a fuzzy neural network model. To show the applicability of the method, it is applied to solve the fuzzy shortest path problem and the fuzzy maximum flow problem. Numerical results illustrate the method accuracy and it’s simple implementation.     

A quantum-implementable neural network model

A quantum-implementable neural network, namely quantum probability neural network (QPNN) model, is proposed in this paper. QPNN can use quantum parallelism to trace all possible network states to improve the result. Due to its unique quantum nature, this model is robust to several quantum noises under certain conditions, which can be efficiently implemented by the qubus quantum computer. Another advantage is that QPNN can be used as memory to retrieve the most relevant data and even to generate new data. The MATLAB experimental results of Iris data classification and MNIST handwriting recognition show that much less neuron resources are required in QPNN to obtain a good result than the classical feedforward neural network. The proposed QPNN model indicates that quantum effects are useful for real-life classification tasks.     

Multimedia data representation learning and applications

Multimedia Tools and Applications -