Logo: University of Southern California

Events Calendar

Select a calendar:

Filter September Events by Event Type:

Events for September 16, 2016

  • Welcome and Welcome Back MFD Students Lunch!

    Fri, Sep 16, 2016 @ 11:00 AM - 12:00 PM

    Mork Family Department of Chemical Engineering and Materials Science

    Student Activity

    Location: EQUAD

    Audiences: MFD STUDENTS ONLY!

    Contact: Martin Olekszyk

    Add to Google CalendarDownload ICS File for OutlookDownload iCal File
  • NL Seminar-Efficient Exploration for Dialog Policy Learning with BBQ Networks & Replay Buffer Spiking

    Fri, Sep 16, 2016 @ 01:30 PM - 02:30 PM

    Information Sciences Institute

    Conferences, Lectures, & Seminars

    Speaker: Zachary Lipton, UCSD

    Talk Title: Efficient Exploration for Dialog Policy Learning with BBQ Networks & Replay Buffer Spiking

    Series: Natural Language Seminar

    Abstract: When rewards are sparse and efficient exploration essential, deep Q learning with e greedy exploration tends to fail. This poses problems for otherwise promising domains such as task oriented dialog systems, where the primary reward signal, indicating successful completion, typically occurs only at the end of each episode but depends on the entire sequence of utterances. A poor agent encounters such successful dialogs rarely, and a random agent may never stumble upon a successful outcome in reasonable time. We present two techniques that significantly improve the efficiency of exploration for deep Q learning agents in dialog systems. First, we demonstrate that exploration by Thompson sampling, using Monte Carlo samples from a Bayes by Backprop neural network, yields marked improvement over standard DQNs with Boltzmann or e greedy exploration. Second, we show that spiking the replay buffer with a small number of successes, as are easy to harvest for dialog tasks, can make Q learning feasible when it might otherwise fail catastrophically.

    Biography: I am a graduate student in the Artificial Intelligence Group at the University of California, San Diego on leave for two quarters at Microsoft Research Redmond. I work on machine learning, focusing on deep learning methods and applications. In particular, I work on modeling sequential data with recurrent neural networks and sequential decision-making processes with deep reinforcement learning. I'm especially interested in research impacting medicine and natural language processing. Recently, in Learning to Diagnose with LSTM RNNs, we trained LSTM RNNs to accurately predict patient diagnoses using only lightly processed time series of sensor readings in the pediatric ICU. Before coming to UCSD, I completed a Bachelor of Arts with a joint major in Mathematics and Economics at Columbia University. Then, I worked in New York City as a jazz musician. I have interned with Amazon's Core Machine Learning team and Microsoft Research's Deep Learning Team.

    Host: Xing Shi and Kevin Knight

    More Info: http://nlg.isi.edu/nl-seminar/

    Location: Information Science Institute (ISI) - 6th Floor -CR # 689; ISI-Marina del Rey

    Audiences: Everyone Is Invited

    Contact: Peter Zamar

    Event Link: http://nlg.isi.edu/nl-seminar/

    Add to Google CalendarDownload ICS File for OutlookDownload iCal File
  • Munushian Visiting Seminar Series

    Munushian Visiting Seminar Series

    Fri, Sep 16, 2016 @ 02:00 PM - 03:30 PM

    Ming Hsieh Department of Electrical and Computer Engineering

    Conferences, Lectures, & Seminars

    Speaker: Dr. Grant Willson, Rashid Engineering Regent Chair, UT Austin

    Talk Title: Polymers for High Resolution Imaging Applications

    Series: Munushian Seminar Series

    Abstract: There has been a continuing and nearly frantic effort on the part of the microelectronics manufacturers over the past
    several decades to make smaller and smaller devices. Companies that cannot keep pace with these advances quickly disappear
    from the market place and sadly many with famous names like Siemens, Motorola and Sony have fallen by the wayside.
    Photolithography, the process that has enabled the production of all of today's microelectronic devices has now reached physical
    limits. Efforts to push that technology to provide still higher resolution by the historical paths of exposure wave length reduction,
    increasing the numerical aperture of the projection lens and reduction in the Raleigh constant have been abandoned. Is this the
    end? Can device scaling continue??
    Various incredibly clever tricks based on chemical engineering principles have been devised that extend the resolution limits of
    photolithography, some of which are already in use in full scale manufacturing. One promising approach for future generations
    of devices is based on the "directed self-assembly" of block co-polymers. We have worked to design block co-polymers that are
    optimized for this application. Doing so requires incorporation of blocks with very high interaction parameters (X) and for some
    applications, incorporation of silicon into one of the blocks. Polymers of this sort form very small structures. We have now
    demonstrated well resolved 50 Angstrom wide lines and spaces. Aligning the structures and orienting them in a way that is useful
    for microelectronics is a challenge as is development of processes for transfer of such small patterns into substrates that are useful
    for device fabrication. A progress report on these efforts will be presented.

    Biography: Dr. Grant Willson is a Professor of Chemical Engineering and Chemistry at the
    University of Texas at Austin where he holds the Rashid Engineering Regent's Chair. He received both his B.S. and Ph.D. in organic chemistry from the University of California, Berkeley and his M.S., in organic chemistry, from San Diego State University. He joined the faculty of the University of Texas at Austin in 1993. Prior to joining the university, Dr. Willson worked at IBM for 17 years as an IBM Fellow and Manager of the Polymer Science and Technology area at the IBM Almaden Research Center in San Jose, CA. He joined IBM after serving on the faculties of California State University, Long Beach and the University of California, San Diego. Dr. Willson is the co-inventor
    of more than 40 issues U.S. patents and co-author of more than 400 publications.

    Dr. Willson's research work is focused on the design and synthesis of functional organic materials with emphasis on organic materials for microelectronics. His work is supported by grants from both government and industry. His research group includes graduate and undergraduate students
    enrolled in both the Chemistry and Chemical Engineering Departments. He was a cofounder of Molecular Imprints, Inc., an Austin firm that employed more than 100 people and was very recently acquired by Canon.

    Host: EE Electrophysics

    Location: Hughes Aircraft Electrical Engineering Center (EEB) - 132

    Audiences: Everyone Is Invited

    Contact: Marilyn Poplawski

    Add to Google CalendarDownload ICS File for OutlookDownload iCal File