Rethinking Stella: comparativa de redes neuronales convolucionales para la clasificación de fulguraciones estelares en curvas de luz de la misión TESS
Autor:
Medina-Baca, Ansony Rolando
Fecha:
01/03/2023Palabra clave:
Tipo de Ítem:
masterThesisResumen:
Este estudio tiene como objetivo mejorar el desempeño del algoritmo Stella en la clasificación de fulguraciones estelares en curvas de luz obtenidas a través de la misión TESS. Para ello, se evaluaron 11 modelos con diferentes arquitecturas de redes neuronales convolucionales (CNN), incluyendo 6 arquitecturas conocidas y 5 nuevas implementaciones. Los resultados demuestran que los 5 modelos propuestos en el segundo experimento mejoran significativamente las métricas de accuracy, precision y recall, logrando una mejora de 2.1%, 3.02% y 7.1% respectivamente en el conjunto de validación. Además, se obtuvo una mejora de 5.4% en la clasificación de verdaderos positivos en el conjunto de validación y 4.42% en el conjunto de test según sus matrices de confusión. Además, las arquitecturas optimizadas permiten reducir en un 67.2% el número de parámetros entrenables con respecto al algoritmo Stella.
Descripción:
This study aims to improve the performance of the Stella algorithm in the classification of stellar flares in light curves obtained through the TESS mission. To do this, 11 models with different convolutional neural network (CNN) architectures were evaluated, including six known architectures and five new implementations. The results show that the five models proposed in the second experiment significantly improve the accuracy, precision, and recall metrics, achieving an improvement of 2.1%, 3.02%, and 7.1% respectively in the validation set. Additionally, a 5.4% improvement was obtained in the classification of true positives in the validation set and 4.42% in the test set according to their confusion matrix. Furthermore, the optimized architectures allow reducing the number of trainable parameters by 67.2% compared to the Stella algorithm.
Ficheros en el ítem
Este ítem aparece en la(s) siguiente(s) colección(es)
Estadísticas de uso
Año |
2012 |
2013 |
2014 |
2015 |
2016 |
2017 |
2018 |
2019 |
2020 |
2021 |
2022 |
2023 |
2024 |
Vistas |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
285 |
224 |
Descargas |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
0 |
174 |
170 |
Ítems relacionados
Mostrando ítems relacionados por Título, autor o materia.
-
Social Voting Techniques: A Comparison of the Methods Used for Explicit Feedback in Recommendation Systems
Nuñez-Valdez, Edward Rolando; Cueva-Lovelle, Juan Manuel; Sanjuan, Oscar; Montenegro-Marin, Carlos Enrique; Infante Hernandez, Guillermo (International Journal of Interactive Multimedia and Artificial Intelligence (IJIMAI), 12/2011)Web recommendation systems usually brings a content list to users based on previous ratings made by them to other similar contents through some social voting mean. This paper aims to present a comparison of the main ... -
Towards an Ontology to Describe the Taxonomy of Common Modules in Learning Management Systems
Montenegro-Marin, Carlos Enrique; Cueva-Lovelle, Juan Manuel; Sanjuan, Oscar; Nuñez-Valdez, Edward Rolando (International Journal of Interactive Multimedia and Artificial Intelligence (IJIMAI), 12/2011)This article have the objective a create ontology for "common modules in a Learning Management Systems", the steps for the build Ontology were: Determine the domain and scope of the ontology, Consider reusing existing ... -
A recommender system based on implicit feedback for selective dissemination of ebooks
Núñez-Valdez, Edward Rolando; Quintana, David; González-Crespo, Rubén ; Isasi, Pedro; Herrera-Viedma, Enrique (Information Sciences, 10/2018)In this study, we describe a recommendation system for electronic books. The approach is based on implicit feedback derived from user's interaction with electronic content. User's behavior is tracked through several ...