Majors
Computer Science and Data Science

Project title
“Function Call Graphs with Code Embeddings from Source Code”

Project description
Alfred developed a tool that enabled future research into program classification source code authorship, plagiarism detection, malware identification, and others. The novel aspect of this tool is the representation of a program as a combination of code2vec embeddings and a call graph. Code2vec is a neural model for representing snippets of code as continuous distributed vectors. Representing a program as a graph allows the relationships between different functions to be preserved which will serve as a better representation of the overall program. 

Faculty Mentor
Dr. Marco Alvarez, Department of Computer Science and Statistics 

---

“I created IR (Intermediate Representation) representations of source code in the form of graphs to be fed into a GNN (Graph Neural Network). We got preliminary results of 70% accuracy for a device mapping task (Determining whether a code kernel will run faster on a GPU or CPU) and are continuing to improve the model to increase performance. Working with Professor Alvarez was great and I am continuing working with him this semester on his research. I hope to continue doing research as an undergraduate and am possibly pursuing research as a career.”