[Colloquium] REMINDER: 10/30 Talks at TTIC: Takahiro Shinozaki, Tokyo Institute of Technology

[Mary Marre via Colloquium]

*When:    *  Tuesday, October 30th at 11:00 am

*Where:     *TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526



*Who:        *Takahiro Shinozaki, Tokyo Institute of Technology


*Title:*       Evolution-Strategy-Based Automation of System Development
for Speech   Recognition and Machine Translation

*Abstract:*
Deep neural networks have been replacing many conventional pattern
recognition systems as well as language processing systems due to its
superior performance. However, they need massive effort by human experts to
tune many meta-parameters specifying the network designs and training
setups.
The hard effort is a prominent obstacle in system development. To automate
the process, we propose to tune the meta-parameters of a whole system using
the evolution strategy and a multi-objective Pareto optimization. We apply
the approach to large vocabulary speech recognition systems and a machine
translation system. Since the method requires repeated training and
evaluation of DNN based systems that require extensive computation, we make
use of parallel computation on cloud computers. Experimental results show
the effectiveness of the proposed approach by discovering appropriate
configuration for the systems automatically. A result of the tuned speech
recognition system is integrated into the Kaldi toolkit as a CSJ recipe for
high-performance Japanese speech recognition.

*Bio:*
Takahiro Shinozaki is an associate professor at Tokyo Institute of
Technology, Tokyo, Japan. He received his B.E., M.E., and Ph.D. degrees in
computer science from the Tokyo Institute of Technology, Tokyo, Japan, in
1999, 2001, and 2004, respectively. From 2004 to 2006, he was a Research
Scholar in the Department of Electrical Engineering, University of
Washington, Seattle. From 2006 to 2007, he was a Research Assistant
Professor in the Academic Center for Computing and Media Studies, Kyoto
University, Kyoto, Japan. From 2007 to 2008 and from 2008 to 2011, he was a
fellowship researcher and an assistant professor in the Department of
Computer Science, Tokyo Institute of Technology, Tokyo, Japan. From 2011 to
2013, he was an assistant professor in the Graduate School of Advanced
Integration Science, Chiba University, Chiba, Japan. His research interests
include semi-supervised and unsupervised learning of spoken languages,
evolutionary algorithms, and their applications. He received the Yamashita
SIG Research Award from Information Processing Society of Japan (IPSJ) in
2009, and the Awaya Prize from the Acoustical Society of Japan (ASJ) in
2008.

*Host: *Kevin Gimpel <kgimpel at ttic.edu>

<kgimpel at ttic.edu>


Mary C. Marre
Administrative Assistant
*Toyota Technological Institute*
*6045 S. Kenwood Avenue*
*Room 517*
*Chicago, IL  60637*
*p:(773) 834-1757*
*f: (773) 357-6970*
*mmarre at ttic.edu <mmarre at ttic.edu>*


On Tue, Oct 23, 2018 at 11:21 PM Mary Marre <mmarre at ttic.edu> wrote:

> *When:    *  Tuesday, October 30th at 11:00 am
>
> *Where:     *TTIC, 6045 S. Kenwood Avenue, 5th Floor, Room 526
>
>
>
> *Who:        *Takahiro Shinozaki, Tokyo Institute of Technology
>
>
> *Title:*       Evolution-Strategy-Based Automation of System Development
> for Speech   Recognition and Machine Translation
>
> *Abstract:*
> Deep neural networks have been replacing many conventional pattern
> recognition systems as well as language processing systems due to its
> superior performance. However, they need massive effort by human experts to
> tune many meta-parameters specifying the network designs and training
> setups.
> The hard effort is a prominent obstacle in system development. To automate
> the process, we propose to tune the meta-parameters of a whole system using
> the evolution strategy and a multi-objective Pareto optimization. We apply
> the approach to large vocabulary speech recognition systems and a machine
> translation system. Since the method requires repeated training and
> evaluation of DNN based systems that require extensive computation, we make
> use of parallel computation on cloud computers. Experimental results show
> the effectiveness of the proposed approach by discovering appropriate
> configuration for the systems automatically. A result of the tuned speech
> recognition system is integrated into the Kaldi toolkit as a CSJ recipe for
> high-performance Japanese speech recognition.
>
> *Bio:*
> Takahiro Shinozaki is an associate professor at Tokyo Institute of
> Technology, Tokyo, Japan. He received his B.E., M.E., and Ph.D. degrees in
> computer science from the Tokyo Institute of Technology, Tokyo, Japan, in
> 1999, 2001, and 2004, respectively. From 2004 to 2006, he was a Research
> Scholar in the Department of Electrical Engineering, University of
> Washington, Seattle. From 2006 to 2007, he was a Research Assistant
> Professor in the Academic Center for Computing and Media Studies, Kyoto
> University, Kyoto, Japan. From 2007 to 2008 and from 2008 to 2011, he was a
> fellowship researcher and an assistant professor in the Department of
> Computer Science, Tokyo Institute of Technology, Tokyo, Japan. From 2011 to
> 2013, he was an assistant professor in the Graduate School of Advanced
> Integration Science, Chiba University, Chiba, Japan. His research interests
> include semi-supervised and unsupervised learning of spoken languages,
> evolutionary algorithms, and their applications. He received the Yamashita
> SIG Research Award from Information Processing Society of Japan (IPSJ) in
> 2009, and the Awaya Prize from the Acoustical Society of Japan (ASJ) in
> 2008.
>
> *Host: *Kevin Gimpel <kgimpel at ttic.edu>
>
> <kgimpel at ttic.edu>
>
>
>
> Mary C. Marre
> Administrative Assistant
> *Toyota Technological Institute*
> *6045 S. Kenwood Avenue*
> *Room 517*
> *Chicago, IL  60637*
> *p:(773) 834-1757*
> *f: (773) 357-6970*
> *mmarre at ttic.edu <mmarre at ttic.edu>*
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20181029/599352c9/attachment.html>