Titre : Keep It Simple: Is Deep Learning Always the Right Tool?

Conférencier : Giulio Antoniol (professeur, DGIGL, Polytechnique Montréal)

Résumé :
Deep neural networks is a popular technique that has been applied successfully to domains such as image processing, sentiment analysis, speech recognition, and computational linguistic. Deep neural networks are machine learning algorithms that, in general, require a labeled set of positive and negative examples that are used to tune hyper-parameters and adjust model coefficients to learn a prediction function. Recently, deep neural networks have also been successfully applied to certain software engineering problem domains (e.g., bug prediction), however, results are shown to be outperformed by traditional machine learning approaches in other domains (e.g., recovering links between entries in a discussion forum).

In this talk, we report our experience in building an automatic Linguistic Antipattern Detector (LAPD) as well as a Self Admitted Technical Debt (SATD)  detector using deep neural networks. We manually build and validate an oracle of around 1,700 instances and create binary classification models using traditional machine learning approaches and Convolutional Neural  Networks.  For the SATD we used a large dataset of manually validated SATD. Our experience is that, considering the size of the oracle, the available hardware and software, as well as the theory to interpret results, deep neural networks are outperformed by traditional machine learning algorithms in terms of all evaluation metrics we used and  resources (time and memory).

Therefore, although deep learning is reported to produce results comparable and even superior to human experts for certain complex tasks, it does not seem to be a good fit for simple classification tasks like smell detection. Researchers and practitioners should be careful when selecting machine learning models for the problem at hand.

Bio :
Giuliano Antoniol is  professor of Software Engineering in the  Department of Computer and  Software  Engineering  of   the  Polytechnique  Montréal  where  he   directs  the  SOCCER laboratory. He  worked in  private companies, research  institutions and  universities. In 2005 he was awarded the Canada Research Chair  Tier I in Software Change and Evolution. He has served in the  program, organization and steering committees of  numerous IEEE and ACM sponsored international conferences and workshops.  His research interest include software traceability,  traceability  recovery  and   maintenance,  software  evolution,  empirical software engineering, search based software engineering, and software testing.

Bienvenue à tous!