World Government and International Equilibrium
Jayantanuja Bandyopadhyaya & Amitava Mukherjee
In continuation of the argument of the preceding chapters for a paradigm shift in IR discourse from synchronic anarchy to a diachronically developing super national regulatory authority within the international system, the salient characteristics and functions of such a global authority are analyzed in this final exercise. A mathematical model is developed, which assumes the existence of an international system consisting of n number of states, and set of three collective goods, namely, security, political stability, and economic welfare, to be distributed among these states. The model shows how a global regulatory authority can ensure the optimal distribution of these collective goods among the constituent units of the international system. The conditions for the success of such a superannuating global authority are also analyzed and mathematically demonstrated. A gradualist approach is adopted, and a case is made out for the democratization of the United Nations and its transformation into a democratic and just world government, in order to ensure the equilibrium and stability of the international system, and the maximization of world peace.
Application of a Radial B asis Function Neural Network for the Diagnosis of Thyroid Disease
Prof. P.N. Chatur & Prof. Dr. A.A. Ghatol
Radial Basis Function (RBF) network describes an artificial neural network model motivated by the "locally tuned" response observed in biological neurons found in many parts of biologic nervous systems. These nervous cells have response characteristics that are selective for some finite range of the initial input signal space. The present model is motivated by earlier work on radial basis function [1] used for interpolation [2], density estimation [3] and approximation of smooth multivariate functions. In this paper, RBF is used for the prediction of thyroid disease. Date from Expert medical practitioner is used for the training and testing of the RBF algorithm for this application.. A comparison of the training times for Back propagation algorithm & Radial basis function Neural network is presented in this paper.
Use of Adaptive Moment um to Reduce Training Time
Dr. Dinesh Kant & N Mahlingam
Neural Networks are being used extensively for the purpose of classification of a signal or data without requiring the mathematical basis for the same. Back propagation (BP) neural network is an easy to implement and simulate and is very robust. After the BPN is trained, it is extremely efficient and fast in classification of the input. But it requires supervision during training. And often BPN are slow to train and convergence rate is not very good.
This paper reports our work where we have been able to reduce the time required for the network to train. The 'delta rule' that is used to determine the modification of the weights in the original BP algorithm has been modified. In the original algorithm, the learning rate has been multiplied with the differential of the error with respect to the weights. This has been modified such that the learning rate is depended on the slope as well as the actual value of the error at that time.. This was achieved by making the learning rate adaptive and introducing the learning rate into the 'learning' algorithm. This has been done to increase the probability of the convergence of the network and not missing a minimum on the error curve.
The network was found to perform extremely well with the modified version of the algorithm. For a specified mean square error the network converged after only 9000 learning cycles compared to 14000 cycles with the original algorithm.
The paper also reports our efforts to study the effect of the size of the network on training and the resultant error during testing. The number of hidden layers and cells (neurons) in each hidden layer were varied while the rate of convergence was measured to determine the efficacy of this technique. It was observed that though the convergence for larger networks was generally faster, increasing the network size beyond a certain size did not improve the performance rate. In fact in certain cases it decreased the performance rate. It is evident that there exists an optimal neural network architecture for every given problem.
Intelligent Single-Agent Control In A Deterministic Environment
F.J. Ogwu, G.A. Aderoummu, E.R. Adagunodo, A.A. Akintola, Adetoye A.O., P.K. Mahanti
An intelligent agent-based approach to control in deterministic environments is hereby presented. A Water Distribution System is used as the test bed. The proposed agent employs the principle of Artificial Intelligence (AI) based on the Artificial Neural Network (ANN) to implement rational behaviour. Several design issues for implementing a neural computational architecture for control of deterministic environments are considered. Object ordinate paradigms are used in modeling the entities in the control system, and a simulation of the system gave results that provide evidence to convergence to perfect agent control.
Multiple Task Allocation With Load Considerations In A Distributed Computing System
Anil Kumar Tripathi, Biplab Kumar Sarker, Naveen Kumar, Deo Prakash Vidyarthi
A Distributed Computing System (DCS) is a network of workstations, personal computers and/or other computing systems. Such a system may be heterogeneous in the sense that the computing nodes may have different speeds and memory capacities. A DCS accepts tasks from users and executes different modules of these tasks on various nodes of the system. Various modules of a task have a precedence relation depicted by its task graph and their communicational requirements are given by the Inter Module Communication (IMC) matrix. A good number of task allocation algorithms have been proposed in the literature. These algorithms allocate a given task on to the DCS nodes and aim to minimize the turn around time of the given task. In this work we have consider (both) the number of modules that can be accepted by the individual computing nodes and the memory capacity of the nodes. Factually, in DCS the nodes may share some specified load within their memory capacity constraints. Further the above mentioned algorithms consider only one given task. In this task we have considered the number of modules that can be accepted by individual nodes along with their memory capacities and arrival of multiple disjoint tasks to the DCS from time to time. We propose an algorithm that will allocate one or more tasks in the DCS when the allocator is invoked. This algorithm will allocate the modules of the tasks, one at a time, that will result in increased throughput of the system as well. Finally, there may be a situation that the given task is not allocated at all due to the high memory requirements of one or more modules of the task.
Forecasting Of Engineering Manpower In Certain Selected Industry Group Using Fuzzy Neural Network Model
J. Paul Choudhury, Dr. Bijan Sarkar, Prof. S.K Mukherjee
Assessment of Engineering Manpower plays a very crucial role to run the Industry. If proper qualified and experienced technical personnel are not available, the Industry cannot run in the most efficient way. Here an effort is made to assess the Engineering Manpower requirement (personnels belonging to Mechanical Engineering) in certain industry group (Steel Manufacturing) in the state of West Bengal for the next five years through a study of selected Engineering Units under particular industry group. The fuzzy Neural Network Method is applied for the assessments of Engineering Manpower requirement.
A Location Chasing Algorithm for Migratory Data Resources On ATM Networks
P.C. Saxena, Sangita Gupta, G. Gabrani
The dynamic migration of data resources such as databases and files has become more popular with the advent of broad-band networks such as ATM. As the data resources migrate from one location to the other, an efficient method is needed to find the location of data resources. In this paper, we propose a new location chasing algorithm with the two existing location management algorithms, by simulation studies under several system parameters such as frequency of remote accesses, frequency of data resource migration and network topology.
Mobile Communication and Computing
Debashis Saha
In this tutorial, we begin with the recent advancements in the realm of wireless and mobile communication technology, which have initiated the paradigm "nomadic computing". Then we discuss the architecture and design concerns that must be addressed as we bring about the system support necessary for nomadicity. Finally, we conclude with the difficult technical issues that separate contemporary systems from the future systems (based on mobile communication and computing) to which we aspire. In brief, the tutorial will cover relevant perspectives relating network, systems, algorithms, and applications that support the symbiosis of portable computers and wireless networks.
An Interlingual Approach to a Cross-Language Information Retrieval System - Case Study for Indian Languages
S. Bandyopadhyay, S. Mukhopadhyay, M. Das
An interlingual approach to a Cross-Language information retrieval system in Indian languages is described. The information will be stored in the form of relations. All operations on the data, i.e., creation, insertion, deletion, updation and query can be specified in any of the Indian language. Transliteration scheme is used to display the data i the language of choice. The input is analyzed using a Template Grammer producing the corresponding SQL statement. The analysis phase works with a number of knowledge bases which are database and language dependent and with a number of dictionaries which are language dependent. The frame structure is used as the interlingual to produce these SQL statements.
Some Observations on Maintainability Metries for Object Oriented Software
Anil Kumar Tripathi, Anil Kumar Malviya
An Object Oriented software system is designed identifying classes and constructing scenarios. Several methodologies including OMT are available for the purpose. Several guidelines are available in literature that suggest various "do's" and "don'ts" to produce an easy to maintain and reliable system. In this work we propose models that capture such ideas and prepare background for development of metrics for assessment of maintainibility of software systems.
A New Buffer Management Scheme for Multi- Qos Traffic Over ATM Switching Systems
G.A. Aderounmu, S.J. Oyeniyi, E.R. Adagunodo and A.D. Akinde
Asynchronous Transfer Mode (ATM) is the technology of choice for the Broadband-Integrated Services Digital Network (B-ISDN). The development of multi-service traffic on the same network requires greater bandwidth and to maintain a good performance of the network so as to guarantee the quality requirements of each class of traffic, several control mechanisms have to be put in place. One such control mechanism in ATM Networks is called Buffer Space Management. In this paper, a novel technique tagged Dynamic Levels of Threshold (DLT), which belongs to Partial Buffer Sharing scheme (PBS) has been proposed for priority buffer management for multi - QoS (Quality of Service) traffics over ATM switching systems. In the proposed model, the amount of both high and low priority cells to be admitted into the buffer in order to prevent loss and congestion is dynamically assigned and varied based on network conditions. The proposed technique is backed up by a numerical analysis for estimating loss probability for both types of cells. An exact analytical model for exact computation of estimating the probability of finding each type of cells through a desired transmission cycles is also used. The performance of the proposed model has been evaluated using a simulation program. The results obtained from studies above show that our model provides for better performance. This new technique can be used for efficient buffer space management and consequently for efficient bandwidth allocation and management aimed at epitomizing network usage in high-speed networks.
Semi-Dynamic Menus versus Static and Dynamic Menus
Musbah Mah'd Aqel, Mahmoud Moh'd Mhashi and P.K. Mahanti
In this paper a different approach to menu selection is proposed. In this approach the users can change the order of sub-menus according to the type of application, user experience, user experience, or skills, and frequency of use or combination leading to a semi-dynamic menu. Two system interfaces are designed and two case studies are performed using the designed system. It is shown that by using the semi-dynamic menu interface, the time performance and the number of keystrokes were reduced by %25 and %31 respectively. However, most users might prefer using a stable menu interface more than others might. This led us to test the user satisfaction of using-dynamic compared with one of the latest dynamic interface called split menus. An experimental split menu interface system was designed and a case study was performed. The result suggest that the subjects are more satisfied and prefer using a semi-dynamic interface rather than dynamic interface.
An Emission Source Control Problem On Air Pollution In Different Classes Of Atmospheric Stability By Finite Difference Mothod
Tripti Chakrabarti, Paritosh Kumar Dhar, Dr. Dilip Kumar Chakraborty
In this paper, an optimal control strategy on air pollution has mathematically been developed taken the emission point source as control such that the area average ground level concentration can be attained in compliance with air quality standard in neutral, stable & unstable atmospheric stability conditions.
Profiles of switching time as upper limit of the control have been obtained on the basis of local available data. The methodology adopted here is the usual maximum principle followed by the use of finite difference method.
|
