High level data classification based on complex network applied to invariant pattern recognition

• Grant number: 13/25876-6
• Support type: Scholarships in Brazil - Doctorate (Direct)
• Effective date (Start): April 01, 2014
• Effective date (End): March 31, 2018
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal researcher: Zhao Liang
• Grantee: Filipe Alves Neto Verri
• Home Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated scholarship(s): 15/18456-6 - Edge dynamics in complex networks for data classification, BE.EP.DD

Abstract

Complex networks have emerged as a unifying topic in complex systems and as a powerful tool of data representation and abstraction since they are able to capture spatial, topological and functional relationship between data. Data classification is a machine learning task that belongs to the supervised learning category. Traditional classification techniques consider only physical features of the input data, i.e. distance, similarity and density. Such techniques are called low-level classification techniques. However, in real cases, data items are not isolated points in the feature space but tend to form certain patterns. Data classification that considers, among to the physical features, the pattern formation is referred as high-level classification. In this project, we aim to develop high-level data classification techniques based on dynamical processes in complex networks. Since high-level classification is able to identify the relationship between the stored patterns and thus extract the pattern formation, it is suitable for solving invariant pattern recognition problems. Thus, the developed methods will be applied in face recognition and handwritten digits and letters recognition. We expect that these methods can recognize images with considerable variation levels, for instance linear variation, illumination and noise, without the need for specific variation treatments. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

NETO VERRI, FILIPE ALVES; URIO, PAULO ROBERTO; ZHAO, LIANG. Network Unfolding Map by Vertex-Edge Dynamics Modeling. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, v. 29, n. 2, p. 405-418, FEB 2018. Web of Science Citations: 2.

Academic Publications

(References retrieved automatically from State of São Paulo Research Institutions)

VERRI, Filipe Alves Neto. Collective dynamics in complex networks for machine learning. 2018. Doctoral Thesis - Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação São Carlos.