search menu icon-carat-right cmu-wordmark

Train but Verify: Towards Practical AI Robustness

Presentation
Train, but Verify is an effort to develop an AI Engineering process to train AI systems to have specific robustness properties.
Publisher

Software Engineering Institute

Watch

Abstract

The project "Train, but Verify" is a multi-year effort to develop an AI Engineering process to train AI systems to have specific robustness properties along with a testing framework capable of verifying these properties. To make this research possible, the SEI created Juneberry, a framework that improves the experience of machine learning experimentation that automates the training, evaluation, and comparison of multiple models against multiple datasets, reducing errors and improving reproducibility.