Renato Porfirio Ishii

Multiclass SVM Model Selection Using Particle Swarm Optimization

Tuning SVM hyperparameters is an important step for achieving good classification performance. In the binary case, the model selection issue is well studied. For multiclass problems, it is harder to choose appropriate values for the base binary models of a decomposition scheme. In this paper, the authors employ Particle Swarm Optimization to perform a multiclass model selection, which optimizes the hyperparameters considering both local and globalmodels. Experiments conducted over 4 benchmark problems show promising results.

An Online Data Access Prediction and Optimization Approach for Distributed Systems

Current scientific applications have been producing large amounts of data. The processing, handling and analysis of such data require large-scale computing infrastructures such as clusters and grids. In this area, studies aim at improving the performance of data-intensive applications by optimizing data accesses. In order to achieve this goal, distributed storage systems have been considering techniques of data replication, migration, distribution, and access parallelism. However, the main drawback of those studies is that they do not take into account application behavior to perform data access optimization. This limitation motivated this paper which applies strategies to support the online prediction of application behavior in order to optimize data access operations on distributed systems, without requiring any information on past executions. In order to accomplish such a goal, this approach organizes application behaviors as time series and, then, analyzes and classifies those series according to their properties. By knowing properties, the approach selects modeling techniques to represent series and perform predictions, which are, later on, used to optimize data access operations. This new approach was implemented and evaluated using the OptorSim simulator, sponsored by the LHC-CERN project and widely employed by the scientific community. Experiments confirm this new approach reduces application execution time in about 50 percent, specially when handling large amounts of data.

Classification of time series generation processes using experimental tools: a survey and proposal of an automatic and systematic approach

By modelling the outputs produced by real world systems, we can study and, therefore, understand how they work and behave under different circumstances. This is especially interesting to support the prediction of future behaviour and, consequently, decision-making, what is particularly required in certain application domains. In order to proceed with such modelling, we organise system outputs as time series and study how those series were generated. The study of the time series generation process typically requires specialists and also detailed information on how and where data was obtained from. However, none of them may be available in certain circumstances. Such limitations motivated this paper which presents a survey of techniques commonly used to evaluate and classify time series generation processes and, most importantly, a novel automatic and systematic approach to conduct such task with a minimum of human intervention and subjectivity. By using such approach, researchers can select adequate techniques to model time series, reducing the modelling time and improving the chances to obtain higher accuracy.

Scheduling based on the impact over process communication of parallel applications

This paper presents a new model for the evaluation of the impacts on processing operations resulting from the communication among processes. The model quantifies the traffic volume imposed on the communication network by means of the latency parameters and the overhead. Such parameters represent the load that each process imposes over the network and the delay on the CPU, as a consequence of the network operations. The delay is represented on the model by means of metric measurements slowdown. The equations that quantify the costs involved in the processing operation and message exchange are defined. In the same way, equations to determine the maximum network bandwidth are used on the decision-making scheduling. The proposed model uses a constant that delimitates the communication network maximum allowed usage, this constant defines two possible scheduling techniques: group scheduling or through communication network. Such techniques are incorporated to the DPWP policy, generating an extension of this policy. Results confirm the performance enhancement of parallel applications.

A complex network-based approach for job scheduling in grid environments

Many optimization techniques have been adopted for efficient job scheduling in grid computing, such as: genetic algorithms, simulated annealing and stochastic methods. Such techniques present common problems related to the use of inaccurate and out-of-date information, which degrade the global system performance. Besides that, they also do not properly model a grid environment. In order to adequately model a real grid environments and approach the scheduling using updated information, this paper uses complex network models and the simulated annealing optimization technique. The complex network concepts are used to better model the grid and extract environment characteristics, such as the degree distribution, the geodesic path, latency. The complex network vertices represent grid process elements, which are generalized as computers. The random and scale free models were implemented in a simulator. These models, associated with Dijkstra algorithm, helps the simulated annealing technique to find out efficient allocation solutions, which minimize the application response time.

Optimizing distributed data access in grid environments by using artificial intelligence techniques

This work evaluates two artificial intelligence techniques for file distribution in Grid environments. These techniques are used to access data on independent servers in parallel, in order to improve the performance and maximize the throughput rate. In this work, genetic algorithms and Hopfield neural networks are the techniques used to solve the problem. Both techniques are evaluated for efficiency and performance. Experiments were conduced in environments composed of 32, 256 and 1024 distributed nodes. The results allow to confirm the decreasing in the file access time and that Hopfield neural network offered the best performance, being possible to be applied on Grid environments.

PheroSLAM: A Collaborative and Bioinspired Multi-agent System Based on Monocular Vision

Multi-robot applications have been extensively discussed and, recently, they are essential for solving problems in robotics field. Nevertheless, development multi-robot real-time applications is usually a complex task, in which it is necessary to design robust environments to support implementation scenarios. In order to deal with such scenarios, this paper proposes PheroSLAM, a bio-inspired multi-robot system based on monocular camera which adopt an extended version of Ant Colony Optimization approach to coordinate multiple-robot teams in the problem related to localization and mapping simultaneously (SLAM). Moreover, robots launch repulsive articial pheromone around themself, creating a repulsive trail in PheroSLAM system. This pheromone trail must be avoided by the other robots, since it denotes an area that have been recently explored. A vision-based SLAM mechanism is also used to provide visual odometry information and to build a 3D feature-based map, considering that every robot must be able to localize itself in the explored environment. Usually, the SLAM problem is solved by cameras or robots remotely controlled. Therefore, the relevance of the proposal is to extend an SLAM problem for many robots and promote the robots move autonomously in the environment according a bio-inspired coordination strategy. Experimental evidences indicated the dispersibility of the PheroSLAM system, increasing the covered area of an environment. Also, results showed that the coordination strategy is efficient and satisfactory to accomplish the exploration task.

IMPROVING SCHEDULING OF COMMUNICATION INTENSIVE PARALLEL APPLICATIONS ON HETEROGENEOUS COMPUTING ENVIRONMENTS

This paper presents a new model for the evaluation of the impacts of processing operations resulting from the communication among processes. This model quantifies the traffic volume imposed on the ...