A comparative analysis of Gabor filters and biologically inspired learning rules for image classification implementing Spiking Neural Networks [recurso electrónico] / Carlos Alan García Campos ; director, Ricardo Morales Carbajal ; codirector, Hirotsugu Okuno
Tipo de material:
TextoDetalles de publicación: Mexicali, Baja California, 2025Descripción: 1 recurso en línea, 69 p. ; il. col., gráficasTema(s): Redes neuronales (Informática) -- Tesis y disertaciones académicas | Redes neurales (computadores) -- Tesis y disertaciones académicas | Redes neuronales (Informática) -- Software -- Tesis y disertaciones académicasClasificación LoC:QA76.87 | G37 2025Recursos en línea: Tesis digital
Nota de disertación: Tesis (Maestría) - - Universidad Autónoma de Baja California, Instituto de Ingeniería, Mexicali, 2025 Resumen: Plastic changes on the synapse drive by spike-timing have been of great interest
as the main learning rule for spiking neural networks. Spike-timing-based rules are
built to model the behavior of a region on the brain related to experimental data in
neuroscience, therefore can lead to differences in the moment of implementing the
rule within a spiking network. This work compares the performance of a pair-wise
and a triplet STDP with different spike interactions to clear an MNIST classification
task. A bio-inspired preprocessing stage was implemented that consisted of a Gabor
filter (as a model of the simple cells mechanism orientation selectivity) and an input
normalization for a homogeneous brightness level of each image. The highlights
of this work are 1) The consistent improvement of the model accuracy whenever
they added the Gabor filter to the inputs; 2) The input normalization to prevent
the overfitting of the model; 3) The Gabor filter helps to correct decoding of some
images of the dataset on the evaluation test.
| Tipo de ítem | Biblioteca actual | Colección | Signatura | Copia número | Estado | Fecha de vencimiento | Código de barras |
|---|---|---|---|---|---|---|---|
| Tesis | Biblioteca Central Mexicali | Colección de Tesis | QA76.87 G37 2025 (Browse shelf(Abre debajo)) | 1 | Disponible | MXL125429 |
Maestría y Doctorado en Ciencias e Ingeniería
Tesis (Maestría) - - Universidad Autónoma de Baja California, Instituto de Ingeniería, Mexicali, 2025
Incluye referencias bibliográficas.
Plastic changes on the synapse drive by spike-timing have been of great interest
as the main learning rule for spiking neural networks. Spike-timing-based rules are
built to model the behavior of a region on the brain related to experimental data in
neuroscience, therefore can lead to differences in the moment of implementing the
rule within a spiking network. This work compares the performance of a pair-wise
and a triplet STDP with different spike interactions to clear an MNIST classification
task. A bio-inspired preprocessing stage was implemented that consisted of a Gabor
filter (as a model of the simple cells mechanism orientation selectivity) and an input
normalization for a homogeneous brightness level of each image. The highlights
of this work are 1) The consistent improvement of the model accuracy whenever
they added the Gabor filter to the inputs; 2) The input normalization to prevent
the overfitting of the model; 3) The Gabor filter helps to correct decoding of some
images of the dataset on the evaluation test.
