TIAN Zengping , WANG Yujun , QU Yunyao , SHI Baile
1997, 8(8):561-568.
Abstract:F-logic language is a logic database language based on frame logic.It is powerful in expressing object-oriented features.However,there was little work in discussing its capability of manipulating complex objects.In this paper,the authors compare the capability of F-logic with that of logic database languages represented by COL (complex object language).Through tWO pairs of semantic preserving transformations,F-logic programs and their Herbrand interpretations can be transformed into COL programs and their corresponding Herbrand interpretations,and vice versa.Also,the effects of negation are discussed.The results of this paper indicate that,without consider-ation of the effects of OID generating,F-logic language has the same power in manipulat-ing complex objects as COL, LDL1(1ogic database language), and ELPS.
1997, 8(8):569-576.
Abstract:This paper presents a representation language based on Rough Set theory,called RSL.This language has tWO parts:one is for application and the other for theory research.The application part is designed mainly for information analysis,such as data analyses and decision making.The research part tries to provide a tool for researchers on theory or on constructing more complicate algorithms.Finding the smallest reduction has been proved tO be an NP-complete problem,a domain-independent approximate algorithm is presented in this paper.It makes the RSL more suitable tO deal with large information tables.
GAO Jingbo , LI Xinyou , TANG Zesheng , ZHOU Xiaohui
1997, 8(8):577-584.
Abstract:This paper introduces a solution to dynamic rectangle intersection searching problem.There are two semi-dynamic algorithms which are based on 1-dimensional data structure and 2-dimensional data structure respectively.The computational complexity Of 1-D searching algorithm is as follows:query time O(1ogM+k'),update time O(1ogMlogn),space O(nlogM).The computational complexity of 2-D searching algorithm is as follows:query time O(1og2M+k),update time O(1og2Mlogn),space O(nlog2M).The two algorithms are implemented respectively.With an experimental comparison,the authors found that 1-D searching algorithm is far better than 2-D searching algorithm.
LAO Zhiqiang , SHI Jiaoying , PAN Yunhe
1997, 8(8):585-592.
Abstract:This paper firstly gives a description of the serial Shading algorithm based on local illumination model, then gives a detailed description of the methods that are used in the distributed Shading algorithm, which are dynamic task allocation method, data adaptive task allocation method, task granularity etc. Later on, the authors give a full description of a new data adaptive algorithm task allocation algorithm, finally give the description of the distributed Shading algorithm and the result.
YI Jianliang , LI Qun , JI Hua , XIE Li
1997, 8(8):593-599.
Abstract:In this paper, the authors propose an adaptive algorithm named ADC, which based on the mathematical theory of evidence, to keep the consistency of shared data in a DSM system. This algorithm can change its policy in keeping data consistency to adapt the behaviors of application programs, so it can add the adaptability of DSM systems greatly.
GAO Yuan , ZHUANG Yan , ZHU Kan , ZHENG Zhenmei , SHI Shugang
1997, 8(8):600-605.
Abstract:The approach presented in this paper uses VM(virtual memory) techniques of OS to allocate a buffer pool from VM whose size can be equal to the size of the whole VM.The approach is implemented as a C+ + class library that can be linked with an application, requiring no special compiler support. Application program accesses objects via normal VM pointers, and manipulates objects using the same compiled code. Dual-buffer strategy based on this VM technique allows the size of database exceed the size of VM.
1997, 8(8):606-609.
Abstract:Snapshot algorithms are fundamental for many distributed applications. This paper shows the several schemes for computing states of channels under the assumption of global clock. Taking consistent cut for the virtual global instant, the authors show that these schemes are suitable for the existing snapshot algorithms. A new algorithm is also presented, which is applicable for the various communication models.
HONG Xiaoguang , WANG Xinjun , DONG Jirun
1997, 8(8):610-614.
Abstract:In this paper, a new hash join algorithm-DHJ(the dynamic hash join) is proposed to resolve the problem of skewed data in the join operation in parallel database.The objective of the algorithm is to avoid the high cost of processing inherent in some early work. Additional buckets are used in the algorithm to balance output during the data partition. Then they are mapped to different processors before the completion of the join operation. The performance analysis of the algorithm is provided in this paper.
WANG Yuguo , PAN Feng , HU Jingqian
1997, 8(8):615-621.
Abstract:In this paper,the authors present an object-oriented framework which adapts to a varietv of dynamic simulation and complex 3D animation controls.The framework is based on a hybrid model.The static structures Of actors are combined as part hierarchies.The attributes of actors determine their shape,appearance,tlme-varying behavior and are defined as class hierarchies.The attributes's dynamic assignment is implemented by dynamic constraint mechanism.This framework has already been used in SimuKit-an inte-grated environment for automatic generation of visual effect for multiple dynamic simula-tion.It is basically implemented on SGl workstation.It is feasible to implement multiple dynamic simulation by this framework.
1997, 8(8):622-629.
Abstract:Binary to binary mapping for classification plays an important role in the researches on feed-forward-neural-network learning.In this paper,the geometrical method is employed to work out a new algorithm to train binary neural networks for classification.By analysis of every training vertex's geometrical location,the algorithm alwavs produces a neural network of four layers for a certain classification problem.The advantages of this algorithm are:it runs with guaranteed convergence and goes to converge much more quickly than BP and some other algorithms;it can determine the structure of the neural networks by learning SO that a precise classification is carried out.In addition,every neuron generated by the algorithm employs a hard-limit activation func-tion with integer synaptic weights,which makes the actual implementation by VLSI tech-nology more facilitated.
CHENG Ling , BAO Zhendong , HUANG Shangteng
1997, 8(8):630-635.
Abstract:In this paper, a kind of multi-lever grammar is introduced. It is the theoretical basis of plugable middleware based on distributed environment. Beginning with the distributed environment based on the client/server architecture, the authors propose the multi-level grammar based on 3-D syntax analysis table, and then expound its supporting mechanism thoroughly, so as to implement the plug and play protocol automation automatically.
1997, 8(8):636-640.
Abstract:The relationship among liberalities of glued theory morphisms and factor theory morphisms in institutions is clarified under certain intuitive conditions, and liberality of the composition of liberal theory morphisms in institutions is shown.