Please use this identifier to cite or link to this item:
http://hdl.handle.net/2445/187143
Title: | A class-conditional approach to learning with noisy labels |
Author: | Catalán Tatjer, Albert |
Director/Tutor: | Radeva, Petia |
Keywords: | Xarxes neuronals (Informàtica) Aprenentatge automàtic Programari Treballs de fi de grau Visió per ordinador Processament digital d'imatges Neural networks (Computer science) Machine learning Computer software Computer vision Digital image processing Bachelor's theses |
Issue Date: | 24-Jan-2022 |
Abstract: | [en] The main topic of this work is Learning with Noisy Labels (LNL). It is the field of Machine Learning concerned with training Neural Networks with noisy datasets. In particular, we have studied DivideMix, a method for LNL in the context of Computer Vision. After an extensive research we have discovered that it is unaware of the underlying class-conditional behaviour which consequently produces class imbalances. In this work, we present two class-conditional approaches to DivideMix. With this intent, we study approximate Baye- sian Inference to quantify per-class uncertainty and leverage this extra information to improve the MixMatch step. In addition, we propose a class-aware policy that improves co-divide. Finally, improving DivideMix’s predictive accuracy by up to 0.39% in certain noise settings. |
Note: | Treballs Finals de Grau d'Enginyeria Informàtica, Facultat de Matemàtiques, Universitat de Barcelona, Any: 2022, Director: Petia Radeva |
URI: | http://hdl.handle.net/2445/187143 |
Appears in Collections: | Treballs Finals de Grau (TFG) - Matemàtiques Programari - Treballs de l'alumnat Treballs Finals de Grau (TFG) - Enginyeria Informàtica |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
codi.zip | Codi font | 516.98 kB | zip | View/Open |
tfg_catalan_tatjer_albert.pdf | Memòria | 3.2 MB | Adobe PDF | View/Open |
This item is licensed under a Creative Commons License