A Visual Display of ML Processes

Proofpoint Computer Science, 2020–21

Liaison(s): Adam Starr, Cameron Malloy
Advisor(s): Elizabeth Sweedyk
Students(s): Juan Diego Herrera (PM), Lindsay Popowski, Alfredo Gomez, Ramya Ramalingam

The Proofpoint team worked on understanding and creating user-friendly explanations for black-box classifiers such as random forests and recurrent neural networks. The resulting explanations have robust theoretical foundations and can be generated efficiently. The team also created appropriate visualizations to complement the explanations so that even non-technical analysts can understand and explain why a black-box classifier arrived at a certain decision.