Select a calendar:
Filter September Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Conferences, Lectures, & Seminars
Events for September
-
Professor Emeritus Michael Arbib: A Remarkable Trajectory - 55 Years of Brains, Machines and Mathematics
Mon, Sep 12, 2016 @ 03:00 PM - 05:30 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Professor Emeritus Michael Arbib, USC
Talk Title: A Remarkable Trajectory - 55 Years of Brains, Machines and Mathematics
Series: CS Keynote Series
Abstract: In honor and celebration of his retirement and 30 years of service at USC, the Viterbi School of Engineering invites Michael A. Arbib to be the inaugural speaker in this series, to share the trajectory of his remarkable career.
To attend, please RSVP by September 5th online at USC.EDU/ESVP (code: arbib). For questions, please contact Cristina Fong, Computer Science Department: 13.821.2981 - cristinf@usc.edu
Biography: The thrust of Michael Arbib's work is expressed in the title of his first book, Brains, Machines and Mathematics (McGraw-Hill, 1964). The brain is not a computer in the current technological sense, but he has based his career on the argument that we can learn much about machines from studying brains, and much about brains from studying machines. He has thus always worked for an interdisciplinary environment in which computer scientists and engineers can talk to neuroscientists and cognitive scientists.
His primary research focus is on the coordination of perception and action. This is tackled at two levels: via schema theory, which is applicable both in top-down analyses of brain function and human cognition as well as in studies of machine vision and robotics; and through the detailed analysis of neural networks, working closely with the experimental findings of neuroscientists on humans and monkeys. He is also engaged in research on the evolution of brain mechanisms for human language, pursuing the Mirror System Hypothesis that links language parity (the fact that what the speaker intends is roughly what the hearer understands) to the properties of the mirror system for grasping -- neurons active for both the execution and observation of actions -- to explain (amongst many other things) why human brains can acquire sign language as readily as speech.
A new interest is working with architects to better understand the neuroscience of the architectural experience and to develop a new field of neuromorphic architecture, "brains for buildings".
The author or editor of almost 40 books, Arbib has most recently edited "Who Needs Emotions? The Brain Meets the Robot" (with Jean-Marc Fellous, Oxford University Press, 2005) and "From Action to Language via the Mirror System" (Cambridge University Press, 2006).
Host: CS Department
Location: Ronald Tutor Hall of Engineering (RTH) - 526
Audiences: Registration Required
Contact: Assistant to CS chair
-
CS Colloquium: Le Song (GATECH) - Discriminative Embedding of Latent Variable Models for Structured Data
Tue, Sep 27, 2016 @ 04:00 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Le Song, GATECH
Talk Title: Discriminative Embedding of Latent Variable Models for Structured Data
Series: Yahoo! Labs Machine Learning Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium. Part of Yahoo! Labs Machine Learning Seminar Series.
Structured data, such as sequences, trees, graphs and hypergraphs, are prevalent in a number of interdisciplinary areas such as network analysis, knowledge engineering, computational biology, drug design and materials science. The availability of large amount of such structured data has posed great challenges for the machine learning community. How to represent such data to capture their similarities or differences? How to learn predictive models from a large amount of such data, and efficiently? How to learn to generate structured data de novo given certain desired properties?
A common approach to tackle these challenges is to first design a similarity measure, called the kernel function, between two data points, based on either statistics of the substructures or probabilistic generative models; and then a machine learning algorithm will optimize a predictive model based on such similarity measure. However, this elegant two-stage approach has difficulty scaling up, and discriminative information is also not exploited during the design of similarity measure.
In this talk, I will present Structure2Vec, an effective and scalable approach for representing structured data based on the idea of embedding latent variable models into a feature space, and learning such feature space using discriminative information. Interestingly, Structure2Vec extracts features by performing a sequence of nested nonlinear operations in a way similar to graphical model inference procedures, such as mean field and belief propagation. In applications involving genome and protein sequences, drug molecules and energy materials, Structure2Vec consistently produces the-state-of-the-art predictive performance. Furthermore, in the materials property prediction problem involving 2.3 million data points, Structure2Vec is able to produces a more accurate model yet being 10,000 times smaller. In the end, I will also discuss potential improvements over current work, possible extensions to network analysis and computer vision, and thoughts on the structured data design problem.
Biography: Le Song is an assistant professor in the Department of Computational Science and Engineering, College of Computing, Georgia Institute of Technology. He received his Ph.D. in Machine Learning from University of Sydney and NICTA in 2008, and then conducted his post-doctoral research in the Department of Machine Learning, Carnegie Mellon University, between 2008 and 2011. Before he joined Georgia Institute of Technology, he was a research scientist at Google. His principal research direction is machine learning, especially kernel methods and probabilistic graphical models for large scale and complex problems, arising from artificial intelligence, network analysis, computational biology and other interdisciplinary domains. He is the recipient of the AISTATS'16 Best Student Paper Award, IPDPS'15 Best Paper Award, NSF CAREER Award'14, NIPS'13 Outstanding Paper Award, and ICML'10 Best Paper Award. He has also served as the area chair or senior program committee for many leading machine learning and AI conferences such as ICML, NIPS, AISTATS and AAAI, and the action editor for JMLR.
Host: Yan Liu
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
-
CS Colloquium and RASC seminar: Ankur Mehta (UCLA) - Pervasive Personal Robots
Thu, Sep 29, 2016 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Ankur Mehta, UCLA
Talk Title: Pervasive Personal Robots
Series: RASC Seminar Series
Abstract: This lecture satisfies requirements for CSCI 591: Computer Science Research Colloquium.
Creating and using new robotic systems has typically been limited to experts, requiring engineering background, expensive tools, and considerable time. Instead, I am working to create systems to automatically design, fabricate, and control functional robots from a simple description of the problem at hand. By enabling the on-demand creation of integrated electromechanical systems by casual everyday users, we can get to a point where we can say for any real-world task, "there's a robot for that."
I have moved towards this vision with a system that can create programmed printable robots from high-level task descriptions. A software-defined-hardware abstraction allows the algorithmic compilation of fabricable subsystem designs from a structural specification; this is in turn generated from a user assisted grounding of a Structured English behavioral specification. The compiled designs are then manufactured using novel printable manufacturing processes, and programmed with autogenerated code. Advanced wireless protocols and communication hardware enable swarms of such robots to interact with each other and users. In this way, fully functional printable robots can be quickly and cheaply designed, fabricated, and controlled to solve custom tasks by casual users.
Biography: Prof. Ankur Mehta is an assistant professor in the Electrical Engineering department of the Henry Samueli School of Engineering and Applied Science at UCLA. Pushing towards his visions of a future filled with robots, his research interests involve printable robotics, rapid design and fabrication, control systems, and wireless sensor networks.
Prof. Mehta was most recently a postdoctoral scholar at MIT's Computer Science and Artificial Intelligence Laboratories investigating design automation for printable robots. Prior to that, he conducted research as a UC Berkeley graduate student in wireless sensor networks and systems, small autonomous aerial robots and rockets, control systems, and micro-elctro-mechanical systems (MEMS).
Prof. Mehta has received best paper awards in the 2015 IEEE Robotics & Automation Magazine and the 2014 International Conference on Intelligent Robots and Systems, and was named a UCLA Samueli Fellow in 2015.
Host: CS Department
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
