Date of Award
Master of Science
Over the past few years, there had been significant achievements in the deployment of deep learning for analysing the programs due to the brilliance of encoding the programs by building vector representations. Deep learning had been used in program analysis for detection of security vulnerabilities using generative adversarial networks, prediction of hidden software defects using software defect datasets. Furthermore, they had also been used for detecting as well as fixing syntax errors that are made by novice programmers by learning a trained neural machine translation on bug-free programming source codes to suggest possible fixes by finding the location of the tokens that are not in place. However, all these approaches either require defect datasets or bug-free code samples that are executable for training the deep learning model. Our neural network model is neither trained with any defect datasets nor bug-free code samples, instead it is trained using structural semantic details of ASTs where each node represents a construct appearing in the source code. This model is implemented to fix one of the most common syntax errors, such as undeclared variable errors as well as infer their type information before program compilation. By this approach, the model has achieved in correctly locating and identifying 81% of the programs on prutor dataset of 1059 programs with undeclared variable errors and also inferring their data types correctly in 80% of the programs.
Venkatesh Theru Mohan
Theru Mohan, Venkatesh, "Automatic repair and type binding of undeclared variables using neural networks" (2019). Graduate Theses and Dissertations. 17582.
Available for download on Sunday, January 12, 2020