Friday, April 16, 2021 – 3:00pm to 4:00pm
Virtual Presentation – ET Remote Access – Rescheduled Event
FILIPE DE AVILA BELBUTE-PERES, Ph.D. Student https://www.linkedin.com/in/filipeabperes
Embedding Physical Knowledge into Deep Learning Models for Efficient and Robust Learning
The recent successes of deep learning methods have motivated their employment to diverse domains, including physical tasks such as predicting the motion of objects or the behavior of fluids. However, in order to learn properly, deep learning methods still require large amounts of data, which can be expensive to obtain in these domains. Moreover, they often do not generalize properly to domains not previously seen in the training data. In this presentation, we show how embedding traditional physics models into deep learning architectures helps to address these issues, allowing for more efficient and robust learning.
Presented in Partial Fulfillment of the CSD Speaking Skills Requirement.
Zoom Participation. See announcement.