<  Back to the Polytechnique Montréal portal

Understanding the Impact of Poor Coding Practices on the Quality of Deep Learning Systems

Hadhemi Jebnoun

Master's thesis (2020)

Open Access document in PolyPublie
[img]
Preview
Open Access to the full text of this document
Terms of Use: All rights reserved
Download (1MB)
Show abstract
Hide abstract

Abstract

Deep Learning (DL) based applications are increasingly being used in our society to solve a variety of tasks, thanks to the recent progress of deep learning models, which are now outperforming humans on a wide range of tasks, from image classification to speech recognition
and natural language processing. This progress is being made towards the widespread application of DL in safety-critical applications such as autonomous cars and healthcare. Deep learning practitioners share similar concerns as software engineers in other domains with regards to efficiency, complexity, and maintainability. On the other hand, the continuous development of deep learning systems which takes place at a rapid pace as well as their increasing complexity could lead to bad design choices on the part of the developers. Furthermore, due to the prevalent use of similar frameworks and repeated coding of similar
tasks, deep learning developers, therefore, tend to use copy-paste practice, creating clones in deep learning code

Résumé

Les applications basées sur l'apprentissage profond, en anglais Deep Learning (DL), sont de plus en plus utilisées pour résoudre diverses tâches de notre quotidien grâce aux recentes prouesses des modèles d'apprentissage profond qui surpassent déjà les compétences humaines dans un large éventail de tâches, de la classification des images à la reconnaissance de la parole et au traitement du langage naturel. Ces progrès tendent à élargir l'application de
l'apprentissage profond dans des domaines aussi critiques en termes de sécurité comme les voitures autonomes et la santé. Les spécialistes de l'apprentissage profond partagent les mêmes préoccupations que les ingénieurs logiciels d' autres domaines en ce qui concerne l'efficacité, la complexité et la maintenabilité des systèmes logiciels. En revanche, le processus de développement continu des systèmes d'apprentissage profond, caractérisé par un rythme rapide et une complexité accrue, pourrait conduire à de mauvais choix de conception par le
développeur.

Department: Department of Computer Engineering and Software Engineering
Program: Génie informatique
Academic/Research Directors: Foutse Khomh
PolyPublie URL: https://publications.polymtl.ca/5542/
Institution: Polytechnique Montréal
Date Deposited: 05 May 2021 12:02
Last Modified: 19 Apr 2023 13:02
Cite in APA 7: Jebnoun, H. (2020). Understanding the Impact of Poor Coding Practices on the Quality of Deep Learning Systems [Master's thesis, Polytechnique Montréal]. PolyPublie. https://publications.polymtl.ca/5542/

Statistics

Total downloads

Downloads per month in the last year

Origin of downloads

Repository Staff Only

View Item View Item