Reuse of a Deep Learning model for handwritten digit recognition

Machine Learning (ML) techniques have made significant advances in solving various problems, which has led to wide dissemination in their use and development. Currently there are different models that have achieved a high level of performance, which raises the question of what to do when we face a p...

Descripción completa

Guardado en:
Detalles Bibliográficos
Autores principales: Pacchiotti, Mauro José, Ballejos, Luciana C., Ale, Mariel Alejandra
Formato: Articulo
Lenguaje:Español
Publicado: 2024
Materias:
Acceso en línea:http://sedici.unlp.edu.ar/handle/10915/168752
Aporte de:
Descripción
Sumario:Machine Learning (ML) techniques have made significant advances in solving various problems, which has led to wide dissemination in their use and development. Currently there are different models that have achieved a high level of performance, which raises the question of what to do when we face a problem for which a very efficient model already exists. This scenario has, for some time, promoted the research and development of different techniques to reuse these models, instead of undertaking the design, implementation, and training of a new one, with all the effort that this entails. In this work, a classification problem is presented, and the reuse of a convolutional neural network is proposed with the objective of recognizing handwritten numbers. Likewise, the performance of the reused model has been evaluated.