Wednesday, November 20, 2019 at 4:00pm to 5:00pm
Building C, 115
113 Research Dr, Bethlehem, PA 18015
Presentation by Sihong Xie, CSE.
Abstract: Structured predictive models based on (neural) graphical models and optimization have state-of-the-art performance in machine learning applications, such as NLP, computational neural science, and fraud detection. However, these models are complex, can be attacked, or trained on noisy data. I am working on the debugging and explanation for structured predictive models, so that the models are transparent, secured, and robust. The methodologies are based on search (in the computer science sense), reinforcement learning, and sensitivity analysis.
NO REGISTRATION REQUIRED.