Logo do repositório
 
Publicação

Single layer optimization with Particle Swarm Optimization: A new approach to optimize Deep Neural Networks

2021-11-09Dissertação de mestrado Acesso aberto
Ver registo simples
dc.contributor.advisorCastelli, Mauro
dc.contributor.advisorBakurov, Illya
dc.contributor.authorMartins, Guilherme de Oliveira Crespo
dc.date.accessioned2021-11-19T12:00:18Z
dc.date.available2021-11-19T12:00:18Z
dc.date.issued2021-11-09
dc.descriptionDissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analyticspt_PT
dc.description.abstractDeep Neural Networks attempt to simulate the behaviour of the brain to solve complex problems.Today, they are currently used for various real world application ssuch as natural language processing, image recognition, self-driving cars,and much more. However, these models, can be very computationally expensive and take a considerable amount of time to train. In this thesis, we attempt to use swarm intelligence to optimize Deep Neural Networks with a smaller computational budget. To achieve this goal, we implement a method that takes any model and selects the layer that can contribute the most for the optimization of said model. Afterwards, we further optimize the layer selected with the Particle Swarm Optimization algorithm in an attempt to take advantage of its ability to surpass local optimums.pt_PT
dc.identifier.tid202792617pt_PT
dc.identifier.urihttp://hdl.handle.net/10362/127958
dc.language.isoengpt_PT
dc.rights.urihttp://creativecommons.org/licenses/by/4.0/pt_PT
dc.subjectMachine learningpt_PT
dc.subjectDeep neural networkspt_PT
dc.subjectParticles warmo ptimizationpt_PT
dc.subjectGradient descentpt_PT
dc.subjectComputer visionpt_PT
dc.titleSingle layer optimization with Particle Swarm Optimization: A new approach to optimize Deep Neural Networkspt_PT
dc.typemaster thesis
dspace.entity.typePublication
rcaap.rightsopenAccesspt_PT
rcaap.typemasterThesispt_PT
thesis.degree.nameMestrado em Métodos Analíticos Avançadospt_PT

Ficheiros

Principais
A mostrar 1 - 1 de 1
Miniatura
Nome:
TAA0107.pdf
Tamanho:
2.68 MB
Formato:
Adobe Portable Document Format
Ver/Abrir

Coleções

NIMS - Dissertações de Mestrado em Ciência de Dados e Métodos Analíticos Avançados (Data Science and Advanced Analytics)