Exploration of Convolutional Neural Network models for source code classification. (January 2021)

• Record Type: Journal Article
• Authors: Barchi, Francesco; Parisi, Emanuele; Urgese, Gianvito; Ficarra, Elisa; Acquaviva, Andrea
• Is Part Of: Engineering applications of artificial intelligence. Volume 97(2021)
• Publication Date: 2021-01
• Subjects: Deep learning -- LLVM-IR -- Code mapping -- Heterogeneous platforms; Engineering -- Data processing -- Periodicals; Artificial intelligence -- Periodicals; Expert systems (Computer science) -- Periodicals; Ingénierie -- Informatique -- Périodiques; Intelligence artificielle -- Périodiques; Systèmes experts (Informatique) -- Périodiques; Artificial intelligence; Engineering -- Data processing; Expert systems (Computer science); Periodicals; 620.00285
• DOI: 10.1016/j.engappai.2020.104075
• Languages: English
• ISSNs: 0952-1976

Abstract

Abstract: The application of Artificial Intelligence is becoming common in many engineering fields. Among them, one of the newest and rapidly evolving is software generation, where AI can be used to automatically optimise the implementation of an algorithm for a given computing platform. In particular, Deep Learning technologies can be used to the decide how to allocate pieces of code to hardware platforms with multiple cores and accelerators, that are common in high performance and edge computing applications. In this work, we explore the use of Convolutional Neural Networks (CNN)s to analyse the application source code and decide the best compute unit to minimise the execution time. We demonstrate that CNN models can be successfully applied to source code classification, providing higher accuracy with consistently reduced learning time with respect to state-of-the-art methods. Moreover, we show the robustness of the method with respect to source code pre-processing, compiler options and hyper-parameters selection. Graphical abstract: Highlights: CNN-based source code classification is effective in terms of accuracy and speed-up. CNNs are more robust than LSTM with respect to hyperparameters and filtering methods. CNNs show a 4-7x shorter training time over LSTM for text and source code analysis. LLVM-IR code analysis is independent of the language and the underlying hardware.