Hyperparameter Learning in Restricted Boltzmann Machines Using Genetic Programming

• Grant number: 18/10706-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: September 01, 2018
• End date: December 31, 2019
• Field of knowledge: Physical Sciences and Mathematics - Computer Science - Computing Methodologies and Techniques
• Principal Investigator: João Paulo Papa
• Grantee: Guilherme Camargo de Oliveira
• Host Institution: Faculdade de Ciências (FC). Universidade Estadual Paulista (UNESP). Campus de Bauru. Bauru , SP, Brazil
• Associated research grant: 14/12236-1 - AnImaLS: Annotation of Images in Large Scale: what can machines and specialists learn from interaction?, AP.TEM
• Associated scholarship(s): 19/13051-9 - Venous leg ulcers assessment using convolutional restricted Boltzmann machines with genetic programming for thermal imaging systems, BE.EP.IC

Abstract

Machine learning techniques have been extensively used in a number of applications, mainly that ones based on deep learning. However, such techniques need to have their hyperparameters fine-tuned for each specific application, being crucial to their good performance. This proposal aims at introducing Genetic Programming (GP) for parameter fine-tuning in Restricted Boltzmann Machines (RBMs), being the results validated in the context of binary image reconstruction. For comparison purposes, other metaheuristic techniques will be considered in the experimental section, as well as other public datasets. As far as we are concerned, GP-based techniques have never been used to fine-tune hyperparameters in RBMs to date. Additonally, the proposal also considers an internship abroad via FAPESP/BEPE.