Deep Semantic Learning Machine Initial design and experiments

Detalhes bibliográficos
Autor(a) principal: Piastka, Konrad
Data de Publicação: 2021
Tipo de documento: Dissertação
Idioma: eng
Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
Texto Completo: http://hdl.handle.net/10362/123465
Resumo: Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics
id RCAP_fd6a08533ff4522396376f391950dc4b
oai_identifier_str oai:run.unl.pt:10362/123465
network_acronym_str RCAP
network_name_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository_id_str 7160
spelling Deep Semantic Learning Machine Initial design and experimentsNeural networksNeuroevolutionGenetic programmingConvolution neural networksImage classificationHeuristic searchNetwork topologiesDissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced AnalyticsComputer vision is an interdisciplinary scientific field that allows the digital world to interact with the real world. It is one of the fastest-growing and most important areas of data science. Applications are endless, given various tasks that can be solved thanks to the advances in the computer vision field. Examples of types of tasks that can be solved thanks to computer vision models are: image analysis, object detection, image transformation, and image generation. Having that many applications is vital for providing models with the best possible performance. Although many years have passed since backpropagation was invented, it is still the most commonly used approach of training neural networks. A satisfactory performance can be achieved with this approach, but is it the best it can get? A fixed topology of a neural network that needs to be defined before any training begins seems to be a significant limitation as the performance of a network is highly dependent on the topology. Since there are no studies that would precisely guide scientists on selecting a proper network structure, the ability to adjust a topology to a problem seems highly promising. Initial ideas of the evolution of neural networks that involve heuristic search methods have provided encouragingly good results for the various reinforcement learning task. This thesis presents the initial experiments on the usage of a similar approach to solve image classification tasks. The new model called Deep Semantic Learning Machine is introduced with a new mutation method specially designed to solve computer vision problems. Deep Semantic Learning Machine allows a topology to evolve from a small network and adjust to a given problem. The initial results are pretty promising, especially in a training dataset. However, in this thesis Deep Semantic Learning Machine was developed only as proof of a concept and further improvements to the approach can be made.Castelli, MauroGonçalves, Ivo Carlos PereiraRUNPiastka, Konrad2021-08-31T14:23:08Z2021-07-202021-07-20T00:00:00Zinfo:eu-repo/semantics/publishedVersioninfo:eu-repo/semantics/masterThesisapplication/pdfhttp://hdl.handle.net/10362/123465TID:202759547enginfo:eu-repo/semantics/openAccessreponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãoinstacron:RCAAP2024-03-11T05:04:48Zoai:run.unl.pt:10362/123465Portal AgregadorONGhttps://www.rcaap.pt/oai/openaireopendoar:71602024-03-20T03:45:03.607025Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informaçãofalse
dc.title.none.fl_str_mv Deep Semantic Learning Machine Initial design and experiments
title Deep Semantic Learning Machine Initial design and experiments
spellingShingle Deep Semantic Learning Machine Initial design and experiments
Piastka, Konrad
Neural networks
Neuroevolution
Genetic programming
Convolution neural networks
Image classification
Heuristic search
Network topologies
title_short Deep Semantic Learning Machine Initial design and experiments
title_full Deep Semantic Learning Machine Initial design and experiments
title_fullStr Deep Semantic Learning Machine Initial design and experiments
title_full_unstemmed Deep Semantic Learning Machine Initial design and experiments
title_sort Deep Semantic Learning Machine Initial design and experiments
author Piastka, Konrad
author_facet Piastka, Konrad
author_role author
dc.contributor.none.fl_str_mv Castelli, Mauro
Gonçalves, Ivo Carlos Pereira
RUN
dc.contributor.author.fl_str_mv Piastka, Konrad
dc.subject.por.fl_str_mv Neural networks
Neuroevolution
Genetic programming
Convolution neural networks
Image classification
Heuristic search
Network topologies
topic Neural networks
Neuroevolution
Genetic programming
Convolution neural networks
Image classification
Heuristic search
Network topologies
description Dissertation presented as the partial requirement for obtaining a Master's degree in Data Science and Advanced Analytics
publishDate 2021
dc.date.none.fl_str_mv 2021-08-31T14:23:08Z
2021-07-20
2021-07-20T00:00:00Z
dc.type.status.fl_str_mv info:eu-repo/semantics/publishedVersion
dc.type.driver.fl_str_mv info:eu-repo/semantics/masterThesis
format masterThesis
status_str publishedVersion
dc.identifier.uri.fl_str_mv http://hdl.handle.net/10362/123465
TID:202759547
url http://hdl.handle.net/10362/123465
identifier_str_mv TID:202759547
dc.language.iso.fl_str_mv eng
language eng
dc.rights.driver.fl_str_mv info:eu-repo/semantics/openAccess
eu_rights_str_mv openAccess
dc.format.none.fl_str_mv application/pdf
dc.source.none.fl_str_mv reponame:Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
instname:Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron:RCAAP
instname_str Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
instacron_str RCAAP
institution RCAAP
reponame_str Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
collection Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
repository.name.fl_str_mv Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos) - Agência para a Sociedade do Conhecimento (UMIC) - FCT - Sociedade da Informação
repository.mail.fl_str_mv
_version_ 1799138056911978496

Registros relacionados