基于模糊神经Petri网的故障诊断模型

Petri网是对具有产生式规则的故障诊断系统的有力建模工具，但其缺乏较强的学习能力．本文以Petri网的基本定义为基础，结合模糊逻辑和Petri网模型，定义了模糊Petri网模型，在此基础上引入人工神经网络技术，给出了人工神经网络的模糊Petri网表示方法，并针对工程机械故障诊断异步、离散等特点，提出并建立了故障诊断的模糊神经Petri网模型及其改进模型．基于模糊神经Petri网的故障诊断系统结合了Petri网和人工神经网络的优点，经过自学习后同时具有很强的推理能力和自适应能力．     

联接主义智能控制综述

综述了近年来联接主义智能控制的理论和应用上的研究进展,覆盖了神经网络的逼近和泛化能力、神经网络与混沌、监督学习算法等基本性质,以及神经网络建模、预测、优化和控制等联接主义智能控制系统的各个部分,并对今后的研究发展提出了展望.     

Context-dependent neural nets-structures and learning

A novel approach toward neural networks modeling is presented in the paper. It is unique in the fact that allows nets' weights to change according to changes of some environmental factors even after completing the learning process. The models of context-dependent (cd) neuron, one- and multilayer feedforward net are presented, with basic learning algorithms and examples of functioning. The Vapnik-Chervonenkis (VC) dimension of a cd neuron is derived, as well as VC dimension of multilayer feedforward nets. Cd nets' properties are discussed and compared with the properties of traditional nets. Possibilities of applications to classification and control problems are also outlined and an example presented.     

Local cluster neural net: Architecture, training and applications

This paper describes the structure, training and computational abilities of the local cluster (LC) artificial neural net architecture. LC nets are a special class of multilayer perceptrons that use sigmoid functions to generate localised functions. LC nets train as fast as radial basis functions nets and are more general. They are well suited for both, multi-dimensional function approximation and discrete classification. The LC net is the result of our search for a widely applicable neural net architecture suitable for low-cost hardware realisation. The LC net seem particularly well suited for analog VLSI realisation of small-size, low-power, fully parallel neural net chip for real time control applications.     

Image recognition with an analog neural net chip

We developed a machine vision system around an analog neural net chip and used it in several applications. Some of them were: locating the address blocks on mail pieces, finding the identification numbers on rail cars, and discriminating between handwritten and machine-printed characters. The chip, operating as a coprocessor of a workstation, provides a speed-up of a factor of 1000, compared with the workstation. The computation speed achieved lies between one and ten billion multiply-accumulates/s. The neural net chip is based on building blocks,neurons, that can be arranged in various network architectures. The dataflow is optimized for implementing large, structured neural nets, and is also suited for any task in which signals are to be convolved with many kernels. Some of the networks are trained on the neural net chip with a weight-perturbation learning algorithm that was adapted to work with the coarse quantization of the weights and the states in the chip.     

粒子群优化算法在神经网络控制中的应用

考虑粒子群优化算法在不确定系统的自适应控制中的应用。神经网络在不确定系统的自适应控制中起着重要作用。但传统的梯度下降法训练神经网络时收敛速度慢,容易陷入局部极小,且对网络的初始权值等参数极为敏感。为了克服这些缺点,提出了一种基于粒子群算法优化的RBF神经网络整定PID的控制策略。首先,根据粒子群算法的基本原理提出了优化得到RBF神经网络输出权、节点中心和节点基宽参数的初值的算法。其次,再利用梯度下降法对控制器参数进一步调节。将传统的神经网络控制与基于粒子群优化的神经网络控制进行了对比,结果表明,后者有更好逼近精度。以PID控制器参数整定为例,对一类非线性控制系统进行了仿真。仿真结果表明基于粒子群优化的神经网络控制具有较强的鲁棒性和自适应能力。     

Genetic design of real-time neural network controllers

The use of genetic algorithms to design neural networks for real-time control of flows in sewerage networks is discussed. In many control applications, standard supervised learning techniques (such as back-propagation) cannot be used through lack of training data. Reinforcement learning techniques, such as genetic algorithms, are a computationally-expensive but viable alternative if a simulator is available for the system in question. The paper briefly describes why genetic algorithms and neural networks were selected, then reports the results of a feasibility study. This demonstrates that the approach does indeed have merits. The implications of high computational cost are discussed, in terms of scaling up to significantly complex problems.     

A Global Asymptotic Stability Result for a Class of Totally Asynchronous Discrete Nonlinear Systems

This paper proves a global stability result for a class of nonlinear discrete-time systems that are subject to regular desynchronization, also known as total asynchronism. The class of systems studied has its origins in a discrete-time neural net model. The techniques used are of interest in terms of the use of a Lyapunov function for the study of convergence of asynchronous nonlinear dynamical systems and also in terms of applications to neural networks. In the latter context, the main result of this paper strengthens a result of an earlier paper on neural networks, and shows that a class of discrete-time continuous-valued neural nets of the Hopfield type displays global convergence properties even when there exists total asynchronism in the updating of neuron states. Date received: May 2, 1995. Date revised: October 19, 1998.     

Fusion of learning automata theory and granular inference systems: ANLAGIS. Applications to pattern recognition and machine learning

In this paper the fusion of artificial neural networks, granular computing and learning automata theory is proposed and we present as a final result ANLAGIS, an adaptive neuron-like network based on learning automata and granular inference systems. ANLAGIS can be applied to both pattern recognition and learning control problems. Another interesting contribution of this paper is the distinction between pre-synaptic and post-synaptic learning in artificial neural networks. To illustrate the capabilities of ANLAGIS some experiments on knowledge discovery in data mining and machine learning are presented. The main, novel contribution of ANLAGIS is the incorporation of Learning Automata Theory within its structure; the paper includes also a novel learning scheme for stochastic learning automata.     

一种自适应模糊Petri网及其应用*

模糊Petri网(fuzzy Petri nets, FPN)是基于模糊产生式规则的知识库系统的有力建模工具,但其缺乏较强的自学习能力。在FPN的基础上引入神经网络技术,给出了一种自适应模糊Petri网(adapt fuzzy Petri nets, AFPN)模型。该模型将神经网络中的BP网络算法引入到FPN模型中,对FPN中的权值进行反复的学习训练,避免了依靠人工经验设置带来的不确定性。AFPN具有很强的推理能力和自适应能力,对知识库系统的建立、更新和维护有着重要的意义。     

A new learning algorithm with logarithmic performance index for complex-valued neural networks

In a fully complex-valued feed-forward network, the convergence of the Complex-valued Back Propagation (CBP) learning algorithm depends on the choice of the activation function, learning sample distribution, minimization criterion, initial weights and the learning rate. The minimization criteria used in the existing versions of CBP learning algorithm in the literature do not approximate the phase of complex-valued output well in function approximation problems. The phase of a complex-valued output is critical in telecommunication and reconstruction and source localization problems in medical imaging applications. In this paper, the issues related to the convergence of complex-valued neural networks are clearly enumerated using a systematic sensitivity study on existing complex-valued neural networks. In addition, we also compare the performance of different types of split complex-valued neural networks. From the observations in the sensitivity analysis, we propose a new CBP learning algorithm with logarithmic performance index for a complex-valued neural network with exponential activation function. The proposed CBP learning algorithm directly minimizes both the magnitude and phase errors and also provides better convergence characteristics. Performance of the proposed scheme is evaluated using two synthetic complex-valued function approximation problems, the complex XOR problem, and a non-minimum phase equalization problem. Also, a comparative analysis on the convergence of the existing fully complex and split complex networks is presented.     

On the Regularity of Petri Net Languages

Petri nets are known to be useful for modeling concurrent systems. Once modeled by a Petri net, the behavior of a concurrent system can be characterized by the set of all executable transition sequences, which in turn can be viewed as a language over an alphabet of symbols corresponding to the transitions of the underlying Petri net. In this paper, we study the language issue of Petri nets from a computational complexity viewpoint. We analyze the complexity of theregularity problem(i.e., the problem of determining whether a given Petri net defines an irregular language or not) for a variety of classes of Petri nets, includingconflict-free,trap-circuit,normal,sinkless,extended trap-circuit,BPP, andgeneralPetri nets. (Extended trap-circuit Petri nets are trap-circuit Petri nets augmented with a specific type ofcircuits.) As it turns out, the complexities for these Petri net classes range from NL (nondeterministic logspace), PTIME (polynomial time), and NP (nondeterministic polynomial time), to EXPSPACE (exponential space). In the process of deriving the complexity results, we develop adecomposition approachwhich, we feel, is interesting in its own right, and might have other applications to the analysis of Petri nets as well. As a by-product, an NP upper bound of the reachability problem for the class of extended trap-circuit Petri nets (which properly contains that of trap-circuit (and hence, conflict-free) and BPP-nets, and is incomparable with that of normal and sinkless Petri nets) is derived.     

基于神经网络实现的自校正PID调节

本文讨论模型未知的离散系统基于神经网络实现的自校正ND调节。本文方法引入能在线评阶控制效果的评价函数，以在线选择神经网络PID调节器的学习样本并决定学习强度，使该调节器能够进行自组织地学习。计算机仿真结果表明，本文方法对模型未知的线性或非线性系统具有良好的调节特性。     

Neurovision-based logic control of an experimental manufacturing plant using neural net le-net5 and automation Petri nets

In this paper, Petri nets and neural networks are used together in the development of an intelligent logic controller for an experimental manufacturing plant to provide the flexibility and intelligence required from this type of dynamic systems. In the experimental setup, among deformed and good parts to be processed, there are four different part types to be recognised and selected. To distinguish the correct part types, a convolutional neural net le-net5 based on-line image recognition system is established. Then, the necessary information to be used within the logic control system is produced by this on-line image recognition system. Using the information about the correct part types and Automation Petri nets, a logic control system is designed. To convert the resulting Automation Petri net model of the controller into the related ladder logic diagram (LLD), the token passing logic (TPL) method is used. Finally, the implementation of the control logic as an LDD for the real time control of the manufacturing system is accomplished by using a commercial programmable logic controller (PLC).     

Complexity Results on Learning by Neural Nets

We consider the computational complexity of learning by neural nets. We are interested in how hard it is to design appropriate neural net architectures and to train neural nets for general and specialized learning tasks. Our main result shows that the training problem for 2-cascade neural nets (which have only two non-input nodes, one of which is hidden) is N-complete, which implies that finding an optimal net (in terms of the number of non-input units) that is consistent with a set of examples is also N-complete. This result also demonstrates a surprising gap between the computational complexities of one-node (perceptron) and two-node neural net training problems, since the perceptron training problem can be solved in polynomial time by linear programming techniques. We conjecture that training a k-cascade neural net, which is a classical threshold network training problem, is also N-complete, for each fixed k2. We also show that the problem of finding an optimal perceptron (in terms of the number of non-zero weights) consistent with a set of training examples is N-hard.Our neural net learning model encapsulates the idea of modular neural nets, which is a popular approach to overcoming the scaling problem in training neural nets. We investigate how much easier the training problem becomes if the class of concepts to be learned is known a priori and the net architecture is allowed to be sufficiently non-optimal. Finally, we classify several neural net optimization problems within the polynomial-time hierarchy.     

Intelligent stock trading system with price trend prediction and reversal recognition using dual-module neural networks

This article presents an intelligent stock trading system that can generate timely stock trading suggestions according to the prediction of short-term trends of price movement using dual-module neural networks(dual net). Retrospective technical indicators extracted from raw price and volume time series data gathered from the market are used as independent variables for neural modeling. Both neural network modules of thedual net learn the correlation between the trends of price movement and the retrospective technical indicators by use of a modified back-propagation learning algorithm. Reinforcing the temporary correlation between the neural weights and the training patterns, dual modules of neural networks are respectively trained on a short-term and a long-term moving-window of training patterns. An adaptive reversal recognition mechanism that can self-tune thresholds for identification of the timing for buying or selling stocks has also been developed in our system. It is shown that the proposeddual net architecture generalizes better than one single-module neural network. According to the features of acceptable rate of returns and consistent quality of trading suggestions shown in the performance evaluation, an intelligent stock trading system with price trend prediction and reversal recognition can be realized using the proposed dual-module neural networks.     

Efficient simulation of hierarchical stochastic activity network models

Stochastic extensions to Petri nets have gained widespread acceptance as a method for describing the dynamic behavior of discrete-event systems. Both simulation and analytic methods have been proposed to solve such models. This paper describes a set of efficient procedures for simulating models that are represented as stochastic activity networks (SANs, a variant of stochastic Petri nets) and composed SAN-based reward models (SBRMs). Composed SBRMs are a hierarchical representation for SANs, in which individual SAN models can be replicated and joined together with other models, in an iterative fashion. The procedures exploit the hierarchical structure and symmetries introduced by the replicate operation in a composed SBRM to reduce the cost of future event list management. The procedures have been implemented as part of a larger performance-dependability modeling package known asUltraSAN, and have been applied to real, large-scale applications. This work was supported in part by the Digital Equipment Corporation Faculty Program: Incentives for Excellence.     

Output feedback adaptive DSC for nonlinear systems with guaranteed L-infinity tracking performance

In this paper, the problem of output tracking for a class of uncertain nonlinear systems is considered. First, neural networks are employed to cope with uncertain nonlinear functions, based on which state estimation is constructed. Then, an output feedback control system is designed by using dynamic surface control (DSC). To guarantee the L-infinity tracking performance, an initialization technique is presented. The main feature of the scheme is that explosion of complex- ity problem in backstepping control is avoided, and there is no need to update the unknown parameters including control gains as well as neural networks weights, the adaptive law with one update parameter is necessary only at the first design step. It is proved that all signals of the closed-loop system are semiglobally uniformly ultimately bounded and the L-infinity performance of system tracking error can be guaranteed. Simulation results demonstrate the effectiveness of the proposed scheme.     

Requirements for intelligent network service operations

A service operation architecture and operation system platform are proposed that separate commonly used information from operations functions, and that use access control functions. This enables new applications to be developed more easily and increases operating efficiency. The operation system platform is related to several surrounding platforms, and requires standardized reference points such as CMIS/P and managed objects. A managed object methodology is a suitable approach for accessing the operation system platform, and managed object classes and methods are proposed for intelligent network service operations. This architecture and platform will allow telecommunication to meet the demands created by intelligent networks for enhanced customer services, more reliable operation systems, and lower development costs. On the basis of proposed platform, service surveillance prototype systems for free-phone services have already been developed, and the next versions of the service operations systems for virtual private networks services are being developed.     

Multi-class pattern classification using neural networks

Multi-class pattern classification has many applications including text document classification, speech recognition, object recognition, etc. Multi-class pattern classification using neural networks is not a trivial extension from two-class neural networks. This paper presents a comprehensive and competitive study in multi-class neural learning with focuses on issues including neural network architecture, encoding schemes, training methodology and training time complexity. Our study includes multi-class pattern classification using either a system of multiple neural networks or a single neural network, and modeling pattern classes using one-against-all, one-against-one, one-against-higher-order, and P-against-Q. We also discuss implementations of these approaches and analyze training time complexity associated with each approach. We evaluate six different neural network system architectures for multi-class pattern classification along the dimensions of imbalanced data, large number of pattern classes, large vs. small training data through experiments conducted on well-known benchmark data.