Center for Data Science and Analytics Talk by Jiatao Gu

RSVP HERE

For non-NYUSH community, please send email to nyush-datascience-group@nyu.edu to RSVP.

Recent research in neural machine translation has largely focused on two aspects: neural network architectures and end-to-end learning algorithms. The problem of decoding, however, has received relatively little attention from the research community. In this talk, we solely focus on the problem of decoding given a trained neural machine translation model. Instead of trying to build a new decoding algorithm for any specific decoding objective, we propose the idea of "trainable decoding" algorithm in which we train a decoding algorithm to find a translation that maximizes an arbitrary decoding objective. More specifically, we design an actor that observes and manipulates the hidden states of the neural machine translation decoder and propose to train it using Reinforcement Learning algorithms. We introduce this idea into two different topics: one is training for real-time neural machine translation, and the other is targeting on trainable greedy decoding in neural machine translation.

Jiatao Gu is currently the 3rd year Ph.D. student in the Department of Electrical and Electronic Engineering at the University of Hong Kong.  He is right now supervised by Prof. Victor O.K. Li.  From June 2016 to January 2017, Jiatao was visiting the Computational Intelligence, Learning, Vision, and Robotics (CILVR), New York University and advised by Dr. Kyunghyun Cho working on neural machine translation and reinforcement learning. Before that, he received his Bachelor's Degree at the Department of Electronic Engineering, Tsinghua University in 2014. His research interests are natural language processing, deep learning and reinforcement learning.

Professor Keith Ross will introduce Jiatao Gu.

This event is sponsored by Center for Data Science and Analytics.