Current Issue : July - September Volume : 2015 Issue Number : 3 Articles : 6 Articles
In real world, many optimization problems are dynamic, which means that their model elements vary with time. These\nproblems have received increasing attention over time, especially from the viewpoint of metaheuristics methods. In this context,\nexperimentation is a crucial task because of the stochastic nature of both algorithms and problems. Currently, there are several\ntechnologies whose methods, problems, and performance measures can be implemented.However, in most of them, certain features\nthat make the experimentation process easy are not present. Examples of such features are the statistical analysis of the results and a\ngraphical user interface (GUI) that allows an easy management of the experimentation process. Bearing in mind these limitations,\nin the present work, we present DynOptLab, a software tool for experimental analysis in dynamic environments. DynOptLab has\ntwo main components: (1) an object-oriented framework to facilitate the implementation of new proposals and (2) a graphical\nuser interface for the experiment management and the statistical analysis of the results. With the aim of verifying the benefits of\nDynOptLab�s main features, a typical case study on experimentation in dynamic environments was carried out....
This paper presents an adaptive memetic algorithm\nto solve the vehicle routing problem with time windows\n(VRPTW). It is a well-known NP-hard discrete optimization\nproblem with two objectivesââ?¬â?to minimize the number\nof vehicles serving a set of geographically dispersed customers,\nand to minimize the total distance traveled in the\nrouting plan. Although memetic algorithms have been proven\nto be extremely efficient in solving the VRPTW, their main\ndrawback is an unclear tuning of their numerous parameters.\nHere, we introduce the adaptive memetic algorithm (AMAVRPTW)\nfor minimizing the total travel distance. In AMAVRPTW,\na population of solutions evolves with time. The\nparameters of the algorithm, including the selection scheme,\npopulation size and the number of child solutions generated\nfor each pair of parents, are adjusted dynamically during the\nsearch. We propose a new adaptive selection scheme to balance\nthe exploration and exploitation of the solution space.\nExtensive experimental study performed on the well-known\nSolomonââ?¬â?¢s and Gehring and Hombergerââ?¬â?¢s benchmark sets\nconfirms the efficacy and convergence capabilities of the proposed\nAMA-VRPTW. We show that it is very competitive\ncompared with other state-of-the-art techniques. Finally, the\ninfluence of the proposed adaptive schemes on the AMAVRPTW\nbehavior and performance is investigated in a thorough\nsensitivity analysis. This analysis is complemented with the two-tailed Wilcoxon test for verifying the statistical significance\nof the results....
A new framework intended for representing and segmenting multidimensional datasets resulting in low spatial complexity\nrequirements and with appropriate access to their contained information is described. Two steps are going to be taken in account.\nThe first step is to specify (n ? 1)D hypervoxelizations, n ? 2, as Orthogonal Polytopes whose nth dimension corresponds to color\nintensity. Then, the nD representation is concisely expressed via the Extreme Vertices Model in the n-Dimensional Space (nDEVM).\nSome examples are presented, which, under our methodology, have storing requirements minor than those demanded by\ntheir original hypervoxelizations. In the second step, 1-Dimensional Kohonen Networks (1D-KNs) are applied in order to segment\ndatasets taking in account their geometrical and topological properties providing a non-supervised way to compact even more the\nproposed n-Dimensional representations.The application of our framework shares compression ratios, for our set of study cases,\nin the range 5.6496 to 32.4311. Summarizing, the contribution combines the power of the nD-EVM and 1D-KNs by producing very\nconcise datasets� representations. We argue that the new representations also provide appropriate segmentations by introducing\nsome error functions such that our 1D-KNs classifications are compared against classifications based only in color intensities.\nAlong the work, main properties and algorithms behind the nD-EVM are introduced for the purpose of interrogating the final\nrepresentations in such a way that it efficiently obtains useful geometrical and topological information....
Class imbalance problem is important issue in design of classifier. This problem results in classifier that gives more errors for minor class. Cost Sensitive approach is efficient to solve this problem in case of neural network. This paper studies effect of cost sensitive learning on Multilayer Perceptron classifier (MLP). The results show that, due to this approach minor class also acquires importance in learning resulting in accuracy of minor class also. Levenberg Marquardt (LM) using new computation (LM) is efficient method for weight updation. But this algorithm can’t be used with complex networks due to large memory requirement. This problem is solved by using new computation technique. Analysis shows that memory requirement has been reduced significantly....
Grid environment is being a service oriented infrastructure in which many heterogeneous resources participate for providing the high performance computation facility to address big problems. Various resources of grid environment exposed as services. So to utilize grid facilities, service discovery becomes an important issue. Some of the services will not be discovered even if they are in same service category due to presence of imprecision and uncertainty in advertised service and requested services in grid environment. We have implemented a rough set algorithm to deal with imprecision and uncertainty. Various services are registered with service registry. This service registry is stored and maintained by Resource Manager Server (RM). Users interact with RM server to find out required service or resource for his job. As these requests may grow large in number, user may need to wait for longer time for response. To deal with such a situation we have implemented a system which creates multiple RM servers that are serving large number of requests in parallel fashion. The evaluation of result shows that implemented system is more efficient than existing one....
We put forward architecture of a framework for integration of data from moving objects related to urban transportation network.\nMost of this research refers to the GPS outdoor geolocation technology and uses distributed cloud infrastructure with big data\nNoSQL database. A network of intelligent mobile sensors, distributed on urban network, produces congestion traffic patterns.\nCongestion predictions are based on extended simulationmodel. Thismodel provides trafficindicators calculations,which fusewith\nthe GPS data for allowing estimation of traffic states across the whole network. The discovery process of congestion patterns uses\nsemantic trajectories metamodel given in our previous works. The challenge of the proposed solution is to store patterns of traffic,\nwhich aims to ensure the surveillance and intelligent real-time control network to reduce congestion and avoid its consequences.\nThe fusion of real-time data from GPS-enabled smartphones integrated with those provided by existing traffic systems improves\ntraffic congestion knowledge, as well as generating new information for a soft operational control and providing intelligent added\nvalue for transportation systems deployment....
Loading....