Select a calendar:
Filter October Events by Event Type:
SUNMONTUEWEDTHUFRISAT
Events for October 15, 2013
-
Phd Defense - Nader Noori
Tue, Oct 15, 2013 @ 02:00 PM - 04:00 PM
Thomas Lord Department of Computer Science
University Calendar
PhD Candidate: Nader Noori
Laurent Itti (Chair)
Michael Arbib
Lisa Azzi-Zadeh (Outside member)
10/15/13
EEB 248
2pm-4pm
Title : The Symbolic Working Memory System
Subtitle: Deriving an Embodied Working Memory Machinery for the Symbolically-Intelligent Mind from Sensorimotor Resources of the Brain
Abstract
Dominant theoretical paradigms for describing the functioning of the brain's short-term memory management systems in the domains of low-level/perception-action and and high-level/intellectual functions follow drastically different principles: embedded and distributed in the low-level domain, disembodied and centralized in the high-level domain. Given that the human cognitive system functions at both levels in different contexts simultaneously this question arises whether indeed there are two types of working memory systems running in parallel under two different operational principles in human brain or, a more parsimonious account can explain all different manifestations of working memory in all domains.
Theoretical inconsistencies and biological/evolutionary implausibility of centralized paradigms of the intellectual domain was a motivation for theorizing about a working memory framework for high-level/intellectual functions based upon control theoretic principles of the low-level functional domain.
The proposed framework demonstrates how novel assemblage of embedded schemas in existing sensorimotor systems may supply a system for management of symbolically represented sensory and motor information serving intellectual tasks. In the
proposed framework, strategic and evolutionarily-constrained reuse of sensorimotor resources for management of respectively spatially-organized and temporally-sensitive information support random access and serial access schemas for management of
symbolic information. Through grounding access schemas for management of symbolic information in sensorimotor systems we are able to predict ramifications of working memory management during the performance of mental tasks at behavioral and neural levels. A detailed example in applying this methodology in well-studied cases of forward and backward recall tasks will be presented with additional computational modeling and the results of simulations.
Our systematic approach in mapping spatial/temporal characteristics of sensorimotor systems onto access modes provides a symbolic interface to other frameworks and
architectures for describing the symbolically-intelligent mind. Proposed framework provides for the first time a neurally-grounded and sensorimotor-based account for management of symbolic information with embodied cognition prospects with opportunities for experimental validations and applications.
In addition to theoretical and computational discussions the result of some experimental studies including eye-tracking during mental sorting tasks will be presented as the supporting evidence for the propose theory.
Location: Hughes Aircraft Electrical Engineering Center (EEB) - 248
Audiences: Everyone Is Invited
Contact: Lizsl De Leon
-
CS Colloquium - Ashutosh Saxena: How should a robot perceive the world?
Tue, Oct 15, 2013 @ 03:30 PM - 05:00 PM
Thomas Lord Department of Computer Science
Conferences, Lectures, & Seminars
Speaker: Ashutosh Saxena, Cornell University
Talk Title: How should a robot perceive the world?
Series: CS Colloquium
Abstract: In order to perform assistive tasks, a robot should learn a functional understanding of the environment. This comprises learning how the objects in the environment could be used (i.e., their affordances). In this talk, I will present methods to represent and learn these affordances using data-driven machine learning algorithms. Our learning algorithm will be Infinite Latent CRFs (ILCRFs) that allow modeling the data with different plausible graph structures. Unlike CRFs where the graph structure is fixed, our ILCRFs learn distributions over possible graph structures in an unsupervised manner.
We then show that our idea of modeling environments using object affordances and (hidden) humans is not only useful for robot manipulation tasks such as arranging a disorganized house, unloading items from a dishwasher, but also in significantly improving standard robotic tasks such as scene segmentation, 3D object detection, human activity detection and anticipation, and task and path planning.
Biography: Ashutosh Saxena is an assistant professor in computer science department at Cornell University. His research interests include machine learning and robotics perception, especially in the domain of robotics in human environments. He received his MS in 2006 and Ph.D. in 2009 from Stanford University, and his B.Tech. in 2004 from Indian Institute of Technology (IIT) Kanpur. He is a recipient of National Talent Scholar award in India, Google Faculty award, Alfred P. Sloan Fellowship, Microsoft Faculty Fellowship, and NSF Career award.
In the past, Ashutosh developed Make3D (http://make3d.cs.cornell.edu), an algorithm that converts a single photograph into a 3D model. Tens of thousands of users used this technology to convert their pictures to 3D. He has also developed algorithms that enable robots (such as STAIR, POLAR, see http://pr.cs.cornell.edu) to perform household chores such as unload items from a dishwasher, place items in a fridge, etc. His work has received substantial amount of attention in popular press, including the front-page of New York Times, BBC, ABC, New Scientist, Discovery Science, and Wired Magazine. He has won best paper awards in 3DRR, IEEE ACE and RSS, and was named a co-chair of the IEEE technical committee on robot learning.
Host: Fei Sha
Location: Henry Salvatori Computer Science Center (SAL) - 101
Audiences: Everyone Is Invited
Contact: Assistant to CS chair
