Grant: Discovery Launch Supplement
University of Alberta
$12,500.00
Finding and Fixing Software Bugs for Increased Data Science Software Reliability and Maintainability
EF 2024-25
Résumé
We seek to innovate software reliability (SR) for data science software systems. SR includes solutions to help developers produce better software, including bug detection and repair solutions that automatically detect whether the application consistently performs its intended functions without errors or failure. SR received much attention in traditional software, but software written by data scientists has yet to benefit from recent advances. We will enable SR for data science by developing threat models and bug taxonomies for data science-specific issues. We will then use such a taxonomy to create extensive open-source datasets of data science bugs and train large code models to detect and repair them. This research is necessary because it targets code written by data scientists, who are often not expert software developers and are, therefore, more likely to write incorrect or buggy code. With the explosion of data science, such programers are numerous and have been under-studied by software engineering research.
SR solutions scan the code written by programmers to detect and highlight any buggy pattern and bad practice, often proposing a bug fix. Such patterns can be manually defined or, following recent advancement of generative AI, learned from millions of examples mined from existing software repositories.
Yet, data science programming systems differ significantly from traditional software and are poorly modeled by existing solutions. Most data science systems rely on interactivity and direct visualizations of results. For example, Jupyter Notebook is a popular framework that differs from existing languages by its interactivity and by combining code, documentation (text and pictures), configuration, and output inside the same document. Such inter-dependency is unusual in traditional software and creates a challenge for existing SR solutions, as many patterns learned from traditional software are no longer applicable.
Statistics Canada estimates that Canadian investment in data science has grown over 400% in the last 15 years, showing the importance of data science software systems. Data science is at the crossroads of many scientific fields, and while experts in their scientific field, data scientists are often laymen regarding software development. As a result, they may inadvertently introduce software bugs and vulnerabilities in their data science software. Hence, software reliability systems that detect and fix bugs are dearly needed but currently fail to address the uniqueness of the data science ecosystem.
Thus, we aim to improve the state of bug detection and repair for data science programs by adapting and inventing new methods and proposing new models to understand, classify, detect, and fix bugs in data science systems. We will introduce software engineering methodologies currently missing in data science to reduce data scientists' struggle with software so that they can focus their efforts on their field of expertise.
Programme
Discovery Launch Supplement
Bénéficiaire
Emplacement
Alberta, Canada
Type de bénéficiaire
Academia
Type
Grant
Source
NSERC's Awards Data
Tel que publié
Application id
DGECR-2024-00347
Name nom
Lutellier, Thibaud
Department département
Augustana Campus - Sciences
Organization id
9
Institution établissement
University of Alberta
Province en
Alberta
Province fr
Alberta
Country en
CANADA
Country fr
CANADA
Fiscal year exercice financier
2024
Competition year année de concours
2024
Award amount
12500
Program id
DGECR
Program name en
Discovery Launch Supplement
Program name fr
Supplément Tremplin vers la découverte
Group en
DISCOVERY RESEARCH
Group fr
RECHERCHE AXÉE SUR LA DÉCOUVERTE
Committee code
1507
Committee name en
Computer Science
Committee name fr
Sciences informatiques
Area of application code
801
Area of application group en
Information and communication services
Area of application group fr
Services d'information et de télécommunications
Area of application en
Computer software
Area of application fr
Logiciel
Research subject code
2706
Research subject group en
Information technology
Research subject group fr
Technologie de l'information
Research subject en
Software engineering
Research subject fr
Génie logiciel
Field of research list names en
Not available
Field of research list names fr
Non disponible
Socioeconomic objective list names en
Not available
Socioeconomic objective list names fr
Non disponible
Application title
Finding and Fixing Software Bugs for Increased Data Science Software Reliability and Maintainability
Keywords
software reliability; data science; defect prediction; automatic program repair; data mining; bug classification
Application summary
We seek to innovate software reliability (SR) for data science software systems. SR includes solutions to help developers produce better software, including bug detection and repair solutions that automatically detect whether the application consistently performs its intended functions without errors or failure. SR received much attention in traditional software, but software written by data scientists has yet to benefit from recent advances. We will enable SR for data science by developing threat models and bug taxonomies for data science-specific issues. We will then use such a taxonomy to create extensive open-source datasets of data science bugs and train large code models to detect and repair them. This research is necessary because it targets code written by data scientists, who are often not expert software developers and are, therefore, more likely to write incorrect or buggy code. With the explosion of data science, such programers are numerous and have been under-studied by software engineering research.
SR solutions scan the code written by programmers to detect and highlight any buggy pattern and bad practice, often proposing a bug fix. Such patterns can be manually defined or, following recent advancement of generative AI, learned from millions of examples mined from existing software repositories.
Yet, data science programming systems differ significantly from traditional software and are poorly modeled by existing solutions. Most data science systems rely on interactivity and direct visualizations of results. For example, Jupyter Notebook is a popular framework that differs from existing languages by its interactivity and by combining code, documentation (text and pictures), configuration, and output inside the same document. Such inter-dependency is unusual in traditional software and creates a challenge for existing SR solutions, as many patterns learned from traditional software are no longer applicable.
Statistics Canada estimates that Canadian investment in data science has grown over 400% in the last 15 years, showing the importance of data science software systems. Data science is at the crossroads of many scientific fields, and while experts in their scientific field, data scientists are often laymen regarding software development. As a result, they may inadvertently introduce software bugs and vulnerabilities in their data science software. Hence, software reliability systems that detect and fix bugs are dearly needed but currently fail to address the uniqueness of the data science ecosystem.
Thus, we aim to improve the state of bug detection and repair for data science programs by adapting and inventing new methods and proposing new models to understand, classify, detect, and fix bugs in data science systems. We will introduce software engineering methodologies currently missing in data science to reduce data scientists' struggle with software so that they can focus their efforts on their field of expertise.