Use of meta-learning for parameter tuning for classification problems

• Grant number: 12/23114-9
• Support Opportunities: Scholarships in Brazil - Doctorate
• Effective date (Start): September 01, 2013
• Effective date (End): May 31, 2018
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: André Carlos Ponce de Leon Ferreira de Carvalho
• Grantee: Rafael Gomes Mantovani
• Host Institution: Instituto de Ciências Matemáticas e de Computação (ICMC). Universidade de São Paulo (USP). São Carlos , SP, Brazil
• Associated scholarship(s): 15/03986-0 - Use of Meta-learning to improve deep learning algorithms in classification problems, BE.EP.DR

Abstract

Currently, a major task in Machine Learning (ML) that has been used is the classification of data, a task that involves assigning a default unknown class of several known ones. In ML, the pattern classification is an instance of supervised learning and can be modeled by a range of algorithms such as Artificial Neural Networks (ANN), Support Vector Machines (SMV), Decision Trees (DT), Deep Learning (DL), among others. Moreover, many times the values of parameters used in such models contribute directly to their performance, and optimize the configuration of these parameters can improve the performance of these algorithms. Recently, meta-learning concepts have been used to choose appropriate settings of parameter sets for algorithms in ML. The use of meta-learning in conjunction with optimization techniques has shown promising results. To handle with the automated selection, we propose in this project to investigate the use of meta-learning both in the selection of algorithms otimization as its parameter sets for classifiers. The overall goal is to achieve better performance in terms of accuracy and computational cost of techniques and algorithms used in classification problems.

Scientific publications

(4)

(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)

MANTOVANI, RAFAEL G.; ROSSI, ANDRE L. D.; ALCOBACA, EDESIO; VANSCHOREN, JOAQUIN; DE CARVALHO, ANDRE C. P. L. F.. A meta-learning recommender system for hyperparameter tuning: Predicting when tuning improves SVM classifiers. INFORMATION SCIENCES, v. 501, p. 193-221, OCT 2019. (12/23114-9, 15/03986-0, 18/14819-5)

AGUIAR, GABRIEL JONAS; MANTOVANI, RAFAEL GOMES; MASTELINI, SAULO M.; DE CARVALHO, ANDRE C. P. F. L.; CAMPOS, GABRIEL F. C.; BARBON JUNIOR, SYLVIO. A meta-learning approach for selecting image segmentation algorithm. PATTERN RECOGNITION LETTERS, v. 128, p. 480-487, DEC 1 2019. (16/18615-0, 12/23114-9, 13/07375-0, 18/07319-6)

CENTINI CAMPOS, GABRIEL FILLIPE; MASTELINI, SAULO MARTIELLO; AGUIAR, GABRIEL JONAS; MANTOVANI, RAFAEL GOMES; DE MELO, LEONIMER FLAVIO; BARBON, JR., SYLVIO. Machine learning hyperparameter selection for Contrast Limited Adaptive Histogram Equalization. EURASIP JOURNAL ON IMAGE AND VIDEO PROCESSING, MAY 6 2019. (12/23114-9)

BARBON, ANA PAULA A. C.; BARBON, JR., SYLVIO; MANTOVANI, RAFAEL GOMES; FUZYI, ESTEFANIA MAYUMI; PERES, LOUISE MANHA; BRIDI, ANA MARIA. Storage time prediction of pork by Computational Intelligence. COMPUTERS AND ELECTRONICS IN AGRICULTURE, v. 127, p. 368-375, SEP 2016. (12/23114-9)

Academic Publications

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

MANTOVANI, Rafael Gomes. Use of meta-learning for hyperparameter tuning of classification problems. 2018. Doctoral Thesis - Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB) São Carlos.