A Machine Learning Based Approach to Detect Machine Learning Design Patterns

A Machine Learning Based Approach to Detect Machine Learning Design Patterns. Pan, W., Washizaki, H., Yoshioka, N., Fukazawa, Y., Khomh, F., & Gu�h�neuc, Y. In Yi, J. & Leavens, G. T., editors, Proceedings of the 30th Asia-Pacific Software Engineering Conference (APSEC), pages 574–578, December, 2023. IEEE CS Press. 5 pages. Early Research Achievements Track.

Paper abstract bibtex

As machine learning expands to various domains, the demand for reusable solutions to similar problems increases. Machine learning design patterns are reusable solutions to design problems of machine learning applications. They can significantly enhance programmers' productivity in programming that requires machine learning algorithms. Given the critical role of machine learning design patterns, the automated detection of them becomes equally vital. However, identifying design patterns can be time-consuming and error-prone. We propose an approach to detect their occurrences in Python files. Our approach uses an Abstract Syntax Tree (AST) of Python files to build a corpus of data and train a refined Text-CNN model to automatically identify machine learning design patterns. We empirically validate our approach by conducting an exploratory study to detect four common machine learning design patterns: Embedding, Multilabel, Feature Cross, and Hashed Feature. We manually label 450 Python code files containing these design patterns from repositories of projects in GitHub. Our approach achieves accuracy values ranging from 80% to 92% for each of the four patterns.

@INPROCEEDINGS{Pan23-APSEC-ERA-ML4MLDPDetection,
   AUTHOR       = {Weitao Pan and Hironori Washizaki and Nobukazu Yoshioka and 
      Yoshiaki Fukazawa and Foutse Khomh and Yann-Ga�l Gu�h�neuc},
   BOOKTITLE    = {Proceedings of the 30<sup>th</sup> Asia-Pacific Software Engineering Conference (APSEC)},
   TITLE        = {A Machine Learning Based Approach to Detect Machine 
      Learning Design Patterns},
   YEAR         = {2023},
   OPTADDRESS   = {},
   OPTCROSSREF  = {},
   EDITOR       = {Joo-yong Yi and Gary T. Leavens},
   MONTH        = {December},
   NOTE         = {5 pages. Early Research Achievements Track.},
   OPTNUMBER    = {},
   OPTORGANIZATION = {},
   PAGES        = {574--578},
   PUBLISHER    = {IEEE CS Press},
   OPTSERIES    = {},
   OPTVOLUME    = {},
   KEYWORDS     = {Topic: <b>Design patterns</b>, Venue: <c>APSEC</c>},
   URL          = {http://www.ptidej.net/publications/documents/APSEC23.doc.pdf},
   PDF          = {http://www.ptidej.net/publications/documents/APSEC23.ppt.pdf},
   ABSTRACT     = {As machine learning expands to various domains, the 
      demand for reusable solutions to similar problems increases. Machine 
      learning design patterns are reusable solutions to design problems of 
      machine learning applications. They can significantly enhance 
      programmers' productivity in programming that requires machine 
      learning algorithms. Given the critical role of machine learning 
      design patterns, the automated detection of them becomes equally 
      vital. However, identifying design patterns can be time-consuming and 
      error-prone. We propose an approach to detect their occurrences in 
      Python files. Our approach uses an Abstract Syntax Tree (AST) of 
      Python files to build a corpus of data and train a refined Text-CNN 
      model to automatically identify machine learning design patterns. We 
      empirically validate our approach by conducting an exploratory study 
      to detect four common machine learning design patterns: Embedding, 
      Multilabel, Feature Cross, and Hashed Feature. We manually label 450 
      Python code files containing these design patterns from repositories 
      of projects in GitHub. Our approach achieves accuracy values ranging 
      from 80\% to 92\% for each of the four patterns.}
}

Downloads: 0

{"_id":"8dBAsMuBBhyC2s6w5","bibbaseid":"pan-washizaki-yoshioka-fukazawa-khomh-guhneuc-amachinelearningbasedapproachtodetectmachinelearningdesignpatterns-2023","author_short":["Pan, W.","Washizaki, H.","Yoshioka, N.","Fukazawa, Y.","Khomh, F.","Gu�h�neuc, Y."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","author":[{"firstnames":["Weitao"],"propositions":[],"lastnames":["Pan"],"suffixes":[]},{"firstnames":["Hironori"],"propositions":[],"lastnames":["Washizaki"],"suffixes":[]},{"firstnames":["Nobukazu"],"propositions":[],"lastnames":["Yoshioka"],"suffixes":[]},{"firstnames":["Yoshiaki"],"propositions":[],"lastnames":["Fukazawa"],"suffixes":[]},{"firstnames":["Foutse"],"propositions":[],"lastnames":["Khomh"],"suffixes":[]},{"firstnames":["Yann-Ga�l"],"propositions":[],"lastnames":["Gu�h�neuc"],"suffixes":[]}],"booktitle":"Proceedings of the 30th Asia-Pacific Software Engineering Conference (APSEC)","title":"A Machine Learning Based Approach to Detect Machine Learning Design Patterns","year":"2023","optaddress":"","optcrossref":"","editor":[{"firstnames":["Joo-yong"],"propositions":[],"lastnames":["Yi"],"suffixes":[]},{"firstnames":["Gary","T."],"propositions":[],"lastnames":["Leavens"],"suffixes":[]}],"month":"December","note":"5 pages. Early Research Achievements Track.","optnumber":"","optorganization":"","pages":"574–578","publisher":"IEEE CS Press","optseries":"","optvolume":"","keywords":"Topic: Design patterns, Venue: <c>APSEC</c>","url":"http://www.ptidej.net/publications/documents/APSEC23.doc.pdf","pdf":"http://www.ptidej.net/publications/documents/APSEC23.ppt.pdf","abstract":"As machine learning expands to various domains, the demand for reusable solutions to similar problems increases. Machine learning design patterns are reusable solutions to design problems of machine learning applications. They can significantly enhance programmers' productivity in programming that requires machine learning algorithms. Given the critical role of machine learning design patterns, the automated detection of them becomes equally vital. However, identifying design patterns can be time-consuming and error-prone. We propose an approach to detect their occurrences in Python files. Our approach uses an Abstract Syntax Tree (AST) of Python files to build a corpus of data and train a refined Text-CNN model to automatically identify machine learning design patterns. We empirically validate our approach by conducting an exploratory study to detect four common machine learning design patterns: Embedding, Multilabel, Feature Cross, and Hashed Feature. We manually label 450 Python code files containing these design patterns from repositories of projects in GitHub. Our approach achieves accuracy values ranging from 80% to 92% for each of the four patterns.","bibtex":"@INPROCEEDINGS{Pan23-APSEC-ERA-ML4MLDPDetection,\r\n AUTHOR = {Weitao Pan and Hironori Washizaki and Nobukazu Yoshioka and \r\n Yoshiaki Fukazawa and Foutse Khomh and Yann-Ga�l Gu�h�neuc},\r\n BOOKTITLE = {Proceedings of the 30th Asia-Pacific Software Engineering Conference (APSEC)},\r\n TITLE = {A Machine Learning Based Approach to Detect Machine \r\n Learning Design Patterns},\r\n YEAR = {2023},\r\n OPTADDRESS = {},\r\n OPTCROSSREF = {},\r\n EDITOR = {Joo-yong Yi and Gary T. Leavens},\r\n MONTH = {December},\r\n NOTE = {5 pages. Early Research Achievements Track.},\r\n OPTNUMBER = {},\r\n OPTORGANIZATION = {},\r\n PAGES = {574--578},\r\n PUBLISHER = {IEEE CS Press},\r\n OPTSERIES = {},\r\n OPTVOLUME = {},\r\n KEYWORDS = {Topic: Design patterns, Venue: <c>APSEC</c>},\r\n URL = {http://www.ptidej.net/publications/documents/APSEC23.doc.pdf},\r\n PDF = {http://www.ptidej.net/publications/documents/APSEC23.ppt.pdf},\r\n ABSTRACT = {As machine learning expands to various domains, the \r\n demand for reusable solutions to similar problems increases. Machine \r\n learning design patterns are reusable solutions to design problems of \r\n machine learning applications. They can significantly enhance \r\n programmers' productivity in programming that requires machine \r\n learning algorithms. Given the critical role of machine learning \r\n design patterns, the automated detection of them becomes equally \r\n vital. However, identifying design patterns can be time-consuming and \r\n error-prone. We propose an approach to detect their occurrences in \r\n Python files. Our approach uses an Abstract Syntax Tree (AST) of \r\n Python files to build a corpus of data and train a refined Text-CNN \r\n model to automatically identify machine learning design patterns. We \r\n empirically validate our approach by conducting an exploratory study \r\n to detect four common machine learning design patterns: Embedding, \r\n Multilabel, Feature Cross, and Hashed Feature. We manually label 450 \r\n Python code files containing these design patterns from repositories \r\n of projects in GitHub. Our approach achieves accuracy values ranging \r\n from 80\\% to 92\\% for each of the four patterns.}\r\n}\r\n\r\n","author_short":["Pan, W.","Washizaki, H.","Yoshioka, N.","Fukazawa, Y.","Khomh, F.","Gu�h�neuc, Y."],"editor_short":["Yi, J.","Leavens, G. T."],"key":"Pan23-APSEC-ERA-ML4MLDPDetection","id":"Pan23-APSEC-ERA-ML4MLDPDetection","bibbaseid":"pan-washizaki-yoshioka-fukazawa-khomh-guhneuc-amachinelearningbasedapproachtodetectmachinelearningdesignpatterns-2023","role":"author","urls":{"Paper":"http://www.ptidej.net/publications/documents/APSEC23.doc.pdf"},"keyword":["Topic: Design patterns","Venue: <c>APSEC</c>"],"metadata":{"authorlinks":{}},"downloads":0},"bibtype":"inproceedings","biburl":"http://www.yann-gael.gueheneuc.net/Work/Publications/Biblio/complete-bibliography.bib","dataSources":["8vn5MSGYWB4fAx9Z4"],"keywords":["topic: design patterns","venue: <c>apsec</c>"],"search_terms":["machine","learning","based","approach","detect","machine","learning","design","patterns","pan","washizaki","yoshioka","fukazawa","khomh","gu�h�neuc"],"title":"A Machine Learning Based Approach to Detect Machine Learning Design Patterns","year":2023}