Use of adaptive techniques in reinforcement learning

• Grant number: 11/17096-5
• Support Opportunities: Scholarships abroad - Research
• Start date: July 30, 2012
• End date: July 29, 2013
• Field of knowledge: Physical Sciences and Mathematics - Computer Science
• Principal Investigator: Ricardo Luis de Azevedo da Rocha
• Grantee: Ricardo Luis de Azevedo da Rocha
• Host Investigator: José Nelson Amaral
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Institution abroad: University of Alberta, Canada

Abstract

This research proposal aims to investigate the use of adaptive techniques in combination with machine learning and to apply this idea to code optimization in compilers. Thus, we seek to improve aspects of code optimization in compilers, an area that has seen relatively recent research efforts, and to expand the use of adaptive techniques in computational intelligence. Learning in compilers has been used to build the scheduling of instructions for basic blocks, to improve on the selection of code transformations that should be applied to a given procedure, to learn a good strategic for loop unrolling, amongst others. Moreover, reinforcement learning is being used in related areas, for example, Ipek et al. proposed a memory controller that can make better decisions about the actions that must perform based on an input queue and also in what the controller has learned from its previous decisions (Ipek 2008). (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)

DA CUNHA RODRIGUES, ELISANGELA SILVA; RODRIGUES, FABRICIO AUGUSTO; DE AZEVEDO DA ROCHA, RICARDO LUIS; CHANG, T. Computational Complexity of Adaptive Algorithms. 2012 THIRD INTERNATIONAL CONFERENCE ON THEORETICAL AND MATHEMATICAL FOUNDATIONS OF COMPUTER SCIENCE (ICTMF 2012), v. 38, p. 6-pg., 2013-01-01. (11/17096-5)