Code Vectors: Understanding Programs Through Embedded Abstracted Symbolic TracesCode Vectors: Understanding Programs Through Embedded Abstracted Symbolic TracesThis research was conducted by Jordan Henkel, Shuvendu K. Lahiri, Ben Liblit, and Thomas Reps. The paper appeared in the 26th ACM SIGSOFT International Symposium on Foundations of Software Engineering (FSE 2018).
With the rise of machine learning, there is a great deal of interest in treating programs as data to be fed to learning algorithms. However, programs do not start off in a form that is immediately amenable to most off-the-shelf learning techniques. Instead, it is necessary to transform the program to a suitable representation before a learning technique can be applied.
In this paper, we use abstractions of traces obtained from symbolic execution of a program as a representation for learning word embeddings. We trained a variety of word embeddings under hundreds of parameterizations, and evaluated each learned embedding on a suite of different tasks. In our evaluation, we obtain 93% top-1 accuracy on a benchmark consisting of over 19,000 API-usage analogies extracted from the Linux kernel. In addition, we show that embeddings learned from (mainly) semantic abstractions provide nearly triple the accuracy of those learned from (mainly) syntactic abstractions.
The full paper is available as a single PDF document. A suggested BibTeX citation record is also available.
We are pleased to offer an experiment artifact containing implementations, artifacts, and other data for replicating and building upon experiments from the paper. The artifact is available from GitHub, and is also available and citable from Zenodo.