End-to-End Learning with Knowledge

Wednesday, March 4, 2020, 11:00 am - 12:00 pm PDTiCal
11th Floor-CR#1135-1137
This event is open to the public.
AI Recruitment Seminar
Bhuwan Dhingra (CMU)
Video Recording:


People increasingly rely on technology to help them navigate the massive amount of online information. Systems which answer user queries must aggregate information from multiple large-scale sources, often in different formats or modalities (e.g. text and databases). In this talk, I will present methods for multi-hop reasoning, which involves following a chain of inferences to derive answers, over a data structure consisting of both structured and unstructured information. By combining classical random-walk algorithms with neural representations these methods require no intermediate supervision beyond question and answer pairs, and scale to the entire Wikipedia in milliseconds. I’ll conclude by laying out future research directions towards a general interface between learning and different types of knowledge.


Bhuwan Dhingra is a final year PhD student at Carnegie Mellon University, advised by William Cohen and Ruslan Salakhutdinov. His research interests are in natural language processing, machine learning and symbolic knowledge representation and reasoning. His work is supported by the Siemens FutureMakers PhD fellowship. In 2019, he won the CMU 3-minute thesis championship. Prior to joining CMU, Bhuwan completed his undergraduate studies at IIT Kanpur in 2013, and spent two years at Qualcomm Research in San Diego.

« Return to Upcoming Events