The dynamics of simple chemical processes is described by multi-layer neural network models. The neural network is able to 'learn' non-linear relationships from the input-out pairs for the process and store th...
详细信息
The dynamics of simple chemical processes is described by multi-layer neural network models. The neural network is able to 'learn' non-linear relationships from the input-out pairs for the process and store the 'knowledge' in paralleldistributedprocessing elements. Neural network models employing back-propagation algorithm may be used to manipulate input variables for the control purpose. The advantage of using such a model for a chemical process is discussed.
based on the assumption that most probability densities in real life can be approximated by a mixture of Gaussian densities, the authors propose a set of algorithms for training a multilayered perceptron as a parallel...
详细信息
ISBN:
(纸本)0879425563
based on the assumption that most probability densities in real life can be approximated by a mixture of Gaussian densities, the authors propose a set of algorithms for training a multilayered perceptron as a paralleldistributedprocessingnetwork (PDP) to estimate various probability densities and serve as a Bayes classifier. The effectiveness of a PDP density estimator was measured in terms of the relative difference between the target probability density function and the network output representing the estimation. The classification rate of the PDP network was effectively identical to that of the Bayes classifier.
A neural-network-based routing algorithm is presented which demonstrates the ability to take into account simultaneously the shortest path and the channel capacity in computer communication networks. A Hopfield-type o...
详细信息
A neural-network-based routing algorithm is presented which demonstrates the ability to take into account simultaneously the shortest path and the channel capacity in computer communication networks. A Hopfield-type of neural-network architecture is proposed to provide the necessary connections and weights, and it is considered as a massively paralleldistributedprocessing system with the ability to reconfigure a route through dynamic learning. This provides an optimum transmission path from the source node to the destination node. The traffic conditions measured throughout the system have been investigated. No congestion occurs in this network because it adjusts to the changes in the status of weights and provides a dynamic response according to the input traffic load. Simulation of a ten-node communication network shows not only the efficiency but also the capability of generating a route if broken links occur or the channels are saturated.
The proceedings contain 32 papers. The special focus in this conference is on Artificial Intelligence. The topics include: A perspective on the nature of artificial intelligence: Enabling and enhancing capabilities fo...
ISBN:
(纸本)9783540520627
The proceedings contain 32 papers. The special focus in this conference is on Artificial Intelligence. The topics include: A perspective on the nature of artificial intelligence: Enabling and enhancing capabilities for society;student modelling in a keyboard scale tutoring system;contradictions and revisions as explanatory aids in the delivery of technical information;a case study in deterministic prolog;exploring the epistemic labyrinth: New directions in the formal theory of knowledge representation;a temporal relational calculus;counterfactuals, cotenability and consistency;a knowledge acquisition tool for decision support systems;adaptive data stores;techniques for efficient empirical induction;representing exceptions in rule-based systems;what can massively parallel architectures bring to AI?;conceptual graphs from a knowledge systems viewpoint;implementing second generation rule-based financial applications today;knowledge in context: A strategy for expert system maintenance;integrating knowledge acquisition and performance systems;a machine vision system with learning capabilities;range from out of focus blur;environment mapping with a mobile robot using sonar;incorporating knowledge via regularization theory: Applications in vision and image processing;common-sense resolution of syntactic ambiguity in database queries;capability based natural language understanding;herbicide advisory systems: Weeds in wheat and other crops;character pattern recognition on a computational neural network;distributed planning and control for manufacturing operations;parallelism in nonmonotonic multiple inheritance systems;a real-time knowledge-based system for frequency management in communications;experiences in developing an intelligent operator guidance system;combining heuristics and simulation models: An expert system for the optimal management of pigs.
Two developing technologies that have become prominent in applied systems architectures are parallelprocessing and distributednetworks. However, the integration of parallel computing devices within commercial and in...
详细信息
Two developing technologies that have become prominent in applied systems architectures are parallelprocessing and distributednetworks. However, the integration of parallel computing devices within commercial and industrial networks is an emerging field with opportunities that have not been sufficiently exploited. One of the reasons for this lag can be found in the conventional thinking about the capabilities of parallelprocessing. Another has been the perceived lack of adequate communication links to handle real-time parallelism over the distances typical for local area networks. This and other implementation difficulties can be overcome quite efficiently through the use of the transputer family of microprocessors with the universal INMOS serial link. With such a transputer based system, one has the advantage of a hardware and software environment that is able to take immediate advantage of the traditional virtues of parallelprocessing at the individual node level, including the ability to expand the number of processors at a node without affecting the performance of other nodes or requiring major changes to the programs operating at the affected node. The net result of such a design, affords the designer and operator of a wide-area network the capabilities of run-time reconfiguration and fault tolerance, combined with relative simplicity in the architecture and in the operational requirements.< >
based on the assumption that most probability densities in real life can be approximated by a mixture of Gaussian densities, the authors propose a set of algorithms for training a multilayered perceptron as a parallel...
详细信息
based on the assumption that most probability densities in real life can be approximated by a mixture of Gaussian densities, the authors propose a set of algorithms for training a multilayered perceptron as a paralleldistributedprocessingnetwork (PDP) to estimate various probability densities and serve as a Bayes classifier. The effectiveness of a PDP density estimator was measured in terms of the relative difference between the target probability density function and the network output representing the estimation. The classification rate of the PDP network was effectively identical to that of the Bayes classifier.< >
A parallel vision architecture called PARADIGM is proposed. PARADIGM has been designed as a mechanism for composing distributed programs for vision processing and spatial-oriented operations. In the PARADIGM environme...
详细信息
A parallel vision architecture called PARADIGM is proposed. PARADIGM has been designed as a mechanism for composing distributed programs for vision processing and spatial-oriented operations. In the PARADIGM environment, users can focus their efforts on solving vision problems and choosing the best strategies for data partitioning without worrying about task allocation/scheduling and details of communications. PARADIGM has been implemented on top of Nectar, a fiber-optic-basednetwork backplane for heterogeneous multicomputers. A prototype spatial database subsystem has been implemented on Nectar using PARADIGM to justify and refine the design of the proposed vision architecture.< >
A parallel algorithm for the implementation of Rete networks on a parallel array processor is proposed. Although special-purpose architectures have been proposed for directly implementing production systems, the avail...
详细信息
A parallel algorithm for the implementation of Rete networks on a parallel array processor is proposed. Although special-purpose architectures have been proposed for directly implementing production systems, the availability of such machines is very limited. Proposed implementation algorithms, which take advantage of more widely available parallel machines, are more practical. Such an approach was taken in a previous work where a number of workstations connected via a local area network were considered. Here an SIMD (single instruction, multiple data) array processor is used. The Rete network is functionally divided into two subnetworks: the discrimination network and the join network. It is the join network where most of the processing load is carried out. The operation of the join network is the main objective of the SIMD implementation. The proposed algorithm emphasizes data parallelism as opposed to functional parallelism. In this type of parallelism, the data are partitioned into sets of approximately equal size and distributed among several processors. The data are then operated upon simultaneously by similar operators. The underlying architecture is described.< >
Decomposition and allocation are common problems in distributed systems. The authors present unified decomposition/allocation solutions to cooperative problem solving; these solutions are expressed in logic programmin...
详细信息
Decomposition and allocation are common problems in distributed systems. The authors present unified decomposition/allocation solutions to cooperative problem solving; these solutions are expressed in logic programming and run on the parallel knowledge-based system. A summary of the parallel knowledge-based system is included. The static allocation schemes described can be realized either automatically or with the advice of a programmer. The schemes maximize the computational resources of a local node and determine the required communication connections of each node within the network. The A* Search program is used to illustrate the allocation schemes. Different decompositions are derived for different schemes.< >
based on the strong analogy between neural networks and distributed diagnosis models, diagnostic algorithms are presented which are similar to the learning algorithm used in neural networks. Diagnostic implications of...
based on the strong analogy between neural networks and distributed diagnosis models, diagnostic algorithms are presented which are similar to the learning algorithm used in neural networks. Diagnostic implications of convergence theorems proved by the Lyapunov function are also discussed. Regarding diagnosis process as a recalling process in the associative memory, a diagnostic method of associative diagnosis is also presented. A good guess of diagnosis is given as a key to recalling the correct diagnosis. The authors regard the distributed diagnosis as an immune network model, a novel PDP (paralleldistributedprocessing) model. This models the recognition capability emergent from cooperative recognition of interconnected units
暂无评论