人工神经网络硬件化途径与神经计算机研究

讨论了人工神经网络硬件化实现的途径和意义，提出对通用神经计算机性能的原则要求．在此原则指引下研制的小型神经计算机（预言神１号），具有灵活性强，使用方便，每秒模拟２×１０７次神经联结计算，并且有在计算过程中随时变更网络拓扑结构和神经元非线性函数的功能，文中讨论了它的适用范围和性能价格比问题．     

Molecular image-processing devices based on chemical reaction systems. I: General principles for implementation

Potentialities for implementing simple neural information-processing devices based on chemical and biochemical dynamical media are discussed. Pilot ‘hardware’ models of neural molecualr devices that able to perform image-processing operations were constructed.     

Molecular image-processing devices based on chemical reaction systems. 4: Image-processing operations performed by active media functioning in the oscillating mode

Basic primitive image-processing operations performed by chemical dynamic media functioning in the oscillating mode are discussed. These operations have rather high computational complexity and can be considered as simulation of human vision capabilities.     

通用高速可变结构并行神经计算机系统性能分析

文中针对通用高速可变结构并行神经计算机系统，在对原系统进行深入分析之后，给出了该系统的时间Ｐｅｔｒｉ网模型，并对其性能进行了分析，给出了分析结果和某些性能指标。     

通用神经网络硬件中神经元基本数学模型的讨论

在介绍了作者实现通用神经网络硬件中应用的通用计算公式的基础上,提出了一种能同时模拟包括RBF与传统BP网络神经元在内的各种神经元通用的新的数学计算模型,并把基于这种通用数学计算模型的神经网络CASSANDRA-Ⅱ型神经计算机结构设计中并予以硬件实现.文中还讨论了它所模拟神经元网络的灵活性.     

Parallel Implementation of Self-Organizing Map on the Partial Tree Shape Neurocomputer

A parallel mapping of self-organizing map (SOM) algorithm is presented for a partial tree shape neurocomputer (PARNEU). PARNEU is a general purpose parallel neurocomputer that is designed for soft computing applications. Practical scalability and a reconfigurable partial tree network are the main architectural features. The presented neuron parallel mapping of SOM with on-line learning illustrates a parallel winner neuron search and a coordinate transfer that are performed in the partial tree network. Phase times are measured to analyse speedup and scalability of the mapping. The performance of the learning phase in SOM with a four processor PARNEU configuration is about 26 MCUPS and the recall phase performs 30 MCPS. Compared to other mappings done for general purpose neurocomputers, PARNEU's performance is very good.     

神经网络计算机的一种分形实现

神经网络计算机的实现是神经网络研究领域中一个重要课题。目前,神经网络的研究已形成了较为系统的理论模型与算法,但神经网络计算机的研究却至今没有重大突破,主要困难就在于网络规模过大,突触联系密度太高等等,为解决这个问题,文中基于分形理论,提出一种神经网络计算机的分形实现方案,给出了分形维数的计算公式,产从物理上实现了与体结构具有自相似性的分形子结构,在神经网络计算机的实现上作了有益的探索。     

柔性电子神经网络计算机结构和原理

本文以开发系统的观点描述了一种新的神经网络计算机体系结构，它以其可开发性和灵活性区别于其他各种神经网络计算机，应用于人工神经网络的研究过程。基于柔性神经网络计算机体系设计，文中提出了一种神经网络分解撕裂算法，为ＶＬＳＩ神经芯片的设计提供了一条新途径。在功能应用上，柔性电子神经网络计算机打破了其他神经网络计算机功能的针对性，为神经网络的研究，开发和应用提供较全面的支持。     

通用数字神经芯片NPS-1的设计和实现

本文深入分析了神经计算的需求，介绍一种通用数字神经芯片ＮＰＳ－１的体系结构和利用ＦＰＧＡ芯片实现样片ＮＰＳ－１。     

On the neurocomputer model of the “living state” of matter and its “biological field” (in the light of G. Ling and E. Bauer’s works)

In this paper, an attempt is made on the basis of the works of G. Ling (USA) and E. Bauer (USSR) to describe and analyze from a unified theoretical and experimental position the neurocomputer model of the “living state” of matter, the physical nature and the mechanisms of physiological resting-state and activity. These are considered as a precondition of understanding anticipatory phenomena in biosystems.