Machine Learning Research at United Technologies Research Center

08 Nov
Wednesday, 11/08/2017 5:00pm to 6:30pm
Computer Science Building, Room 151
Career Development

Dr. Hala Mostafa and Dr. Tong Sun will present a broad overview of UTRC's Systems Department research in machine learning, computer vision and human machine interaction.

Then they will give a deep dive on the topic of Transfer Learning with Expert Knowledge. At UTRC, they work with the business units to build predictive models from historical data. Despite the furor around Big Data, in many settings obtaining more data continues to be expensive or impractical. Transfer learning (TL) tries to leverage data-rich source tasks to improve learning on different but related target tasks. But TL approaches typically require lots of unlabeled target data, which again may be impractical. Domain experts (e.g., engineers, finance people) usually have some knowledge of the anticipated changes (AC) between the source and target tasks (e.g., in a price prediction task, the expert may know the maximum price increase for certain items). They discuss how they leveraged ACs to improve the data-efficiency of TL and the assumptions they made to do so. They will also touch on some of the domain adaptation work to address more general TL settings.

They will conclude with research problems of interest to UTRC and discuss existing and future career and internship opportunities in the broad area of computer vision and machine learning.

 

Space is limited. Register now on Handshake!

If you do not have a Handshake account, email us.