Machine learning in complex networks

Author(s)

Fabricio Aparecido Breve

• Document type: Doctoral Thesis
• Press: São Carlos.
• Institution: Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB)
• Defense date: 2010-08-23

Examining board members

Zhao Liang; Jose Roberto Castilho Piqueira; Roseli Aparecida Francelin Romero; Nei Yoshihiro Soma; Renato Tinós

• Advisor: Zhao Liang

Abstract

Complex networks is a recent and active scientific research field, which concerns large scale networks with non-trivial topological structure, such as computer networks, telecommunication networks, transport networks, social networks and biological networks. Many of these networks are naturally divided into communities or modules and, therefore, uncovering their structure is one of the main problems related to complex networks study. This problem is related with the machine learning field, which is concerned with the design and development of algorithms and techniques which allow computers to learn, or increase their performance based on experience. Some of the problems identified in traditional learning techniques include: difficulties in identifying irregular forms in the attributes space; uncovering overlap structures of groups or classes, which occurs when elements belong to more than one group or class; and the high computational complexity of some models, which prevents their application in larger data bases. In this work, we deal with these problems through the development of new machine learning models using complex networks and space-temporal dynamics. The developed models have performance similar to those from some state-of-the-art algorithms, at the same time that they present lower computational complexity order than most of them (AU)

• FAPESP's process: 07/00222-2 - Devolopement of a New Data Clustering Technique Using Dynamic Model and Active Contours
• Grantee: Fabricio Aparecido Breve
• Support type: Scholarships in Brazil - Doctorate