Data classification based on random walk on networks

• Grant number: 12/14217-9
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): September 01, 2012
• Effective date (End): January 31, 2014
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: Zhao Liang
• Grantee: Filipe Alves Neto Verri
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil

Abstract

Through computational representation, the machine learning techniques can generate models capable of organizing the existing knowledge or mimicking the behavior of a human expert in the relevant fields. Traditional classification techniques only consider physical features of the data such as distance, similarity and density. These classifiers are called low-level classifiers. Often the observation of these features is not sufficient to classify correctly the data. Other current techniques also consider the features of pattern formation, which have semantic meanings. These techniques are called high-level classifiers. The high-level classification solves many problems such as invariant pattern recognition. In this project, it is proposed to develop a high-level classifier based on networks by using measures related to the random walk. This classifier should be able to look beyond the physical features of the data, the features of the pattern formation to which they belong.(AU)