Computational Systems

Summary

 

Historically, PEC’s interest in computer science began in the seventies with the development of numerical methods for structure analysis. Increasingly complex problems involving, for example, soil-fluid-structure interactions, and the uncertainties inherent to natural phenomena enlarged the scope of the researches in the area of structures, promoting the need for other computational resources. In the eighties the development of parallel and vector computers allowed to solve more complex problems, involving a great number of variables. Computational methods, so far limited to deterministic solutions, started working with uncertainties. From the nineties, the evolution of algorithms and the increasing capacity of computers allowed to develop robust models from databases. Recently, research has been focused on the extraction of knowledge from large document collections.

Researches in the Computational Systems Area are concentrated on the development of computational models, using computational resources in parallel distributed environments, databases, graphic views, etc. Development of Computational Intelligence and Data Mining techniques for complex system modeling have been the main focus of researches. The work mainly involves the development of algorithms based on fuzzy logic, neural networks, support vector machines, genetic algorithms and nature inspired optimization techniques. Those techniques are widely applied in most knowledge areas allowing the development of partnership projects with other PEC research areas as well as other programs from COPPE.

                              

The work in this area has been developed through institutional and technological research projects focusing the development of complex system models for several applications in oil industry, environment, energy, business and others. Applications are not limited to the Engineering field, and the area also acts in bioinformatics problems, remote sensing, and ecology among others, always in partnership with specialists of those areas.

                                               

The impact of the internet on human development has been compared to the invention of the printing press with deep technological, economic and social transformations promoted by information scattering on the web. In this context, the development of scientific research in the Area of Computational Systems has aimed at knowledge extraction from document collections and methodologies to increase knowledge generation and the network cooperation.

                                                                   

Several researches are fulfilled in partnership with other programs from COPPE. The Technology Transfer Nucleus (NTT) is the laboratory at PEC that acts directly in this area and has a strong interaction with the Laboratory for Computational Methods in Engineering (LAMCE) and the High Performance Computing Nucleus (NACAD) from COPPE.

                                                           

Research Lines

Scientific Visualization & Virtual Reality

Development of systems aiming the visualization of numerical modeling process and data mining results.

Complexity and Knowledge

This research line focuses several aspects of the complexity of technological, biological and social systems. It approaches knowledge processes on certain system learning and problem solution. Uses fractals in physical phenomena, not limitted to geometric representations. Develops models for knowledge transference. Research projects aim the integration of several technologies.

Optimization and Computational Methods Inspired in Nature

This line seeks the development of optimization algorithms based on classical methods (mathematical programming) and in nature inspired methods such as genetic algorithms, swarm intelligence, and artificial immunological systems among others. Researches in this line aim to improve algorithms for several applications mainly in offshore engineering and structure projects.

Data Models  & Knowledge

Development of new computational intelligence algorithms for data mining for different applications. This line deals with data models useful for knowledge extraction in complex application in engineering, bioinformatics, business and others.

Analysis of Non-structured Information

Development of algorithms and systems for text and web mining. Projects in this line approach all the steps in text mining process: pre-processing, data mining algorithm adaptation to text mining, visualization, knowledge discovery on the web (web mining: surf, link content and analysis).

 

Big Data

The main focus is the aggregation and processing of large data sets obtained in real time in various repositories and sensors, plugged into the web or equipment, that continuously capture structured and unstructured information in order to generate knowledge, considering the criteria of efficiency and privacy. There are various fields and activities that need to analyze information from large volumes of data such as: oil and gas, biology, environment, meteorology, satellite images, social networks, competitive intelligence, security, financial markets, e-commerce, etc…

Decision Making, Analysis of Uncertainties and Risks

This line researches methods for uncertainty and risk analysis on several applications. Projects on this research line comprise the development of uncertainty analysis test methods applied to financial, oil and ecology sectors and decision making processes.

Complex Network Computational Modeling

The theory of complex networks and the mathematical formalism of graph theory are approached in this line. New algorithms in computational intelligence are developed for several applications.  Research projects aim the integration of modeling technologies of complex systems.

 

Human Mobility Patterns

The research uses data from cell phone with the record of each link is associated with geographical coordinates of the antenna that has processed the call. Studies have shown that human displacement has a high degree of spatial and temporal regularity. After correction for differences in the distances and anisotropy inherent in each trajectory, the displacement patterns converge to the same probability distribution, indicating that despite the great diversity of trajectories, human mobility follows reproducible patterns. Thus one can build efficient models of urban dynamics in the areas of transportation, spreading viruses, epidemics, land use, and so on...