[Colloquium] CS Seminar today at 3:30 pm: Bo Dai, Georgia Institute of Technology
Sandra Wallace via Colloquium
colloquium at mailman.cs.uchicago.edu
Tue Mar 13 13:14:22 CDT 2018
UNIVERSITY OF CHICAGO
DEPARTMENT OF COMPUTER SCIENCE
PRESENTS
Bo Dai
Georgia Institute of Technology
Tuesday, March 13, 2018 at 3:30 pm
Ryerson 251
Title: Exploiting the Recursive Structure in Machine Learning
Abstract:
Machine learning has recently witnessed revolutionary success in a wide spectrum of domains. Most of these applications involve learning with complex inputs, e.g., the infinite horizon sequences in reinforcement learning and the graphs in chemical and material design. The success of these applications of machine learning techniques often requires at least two factors: i) the exploitation of structure information in learning models, and ii) the utilization of huge amount of data. However, the structure information corresponds delicate conditions in optimization point of view, while a huge amount of data requires algorithms efficient and scalable. Integrating both parts can be very challenging, from both computational and theoretical perspectives.
In this talk, I will share my research efforts on developing principled, scalable and practical algorithms and models for learning with the recursive structures. Specifically, I will discuss our reinforcement learning algorithm which exploits the recursive structure in Bellman optimality equation. This work takes a substantial step towards solving the decades-long open problem in reinforcement learning for seeking a convergent algorithm with function approximations on off-policy data. I will also present our work, ‘structure2vec’, which exploits the recursive structure in an alternative way for handling graph inputs. Empirical results show the structure2vec achieves the state-of-the-art accuracy with smaller model size while faster training speed.
Bio:
Bo Dai is a Ph.D. candidate in Computer Science at Georgia Institute of Technology. His research interests lie in developing effective statistical models and efficient algorithms for learning from a massive volume of complex and structured data, including large-scale optimization, reinforcement learning, and structured data modeling. He is the recipient of the best paper award of AISTATS2016 and NIPS2017 workshop on Machine Learning for Molecules and Materials.
Host: Risi Kondor
Refreshments served after the talk in Ry. 255
Link to PDF: https://www.cs.uchicago.edu/sites/cs/files/uploads/seminar_announcements/Dai%20Poster.pdf <https://www.cs.uchicago.edu/sites/cs/files/uploads/seminar_announcements/Dai%20Poster.pdf>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20180313/54ee4cdc/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PastedGraphic-2.png
Type: image/png
Size: 29965 bytes
Desc: not available
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20180313/54ee4cdc/attachment-0001.png>
More information about the Colloquium
mailing list