[Colloquium] Reminder: De la Torre Talk Tomorrow at 1:30pm TTI-C

[Katherine Cumming]

TOYOTA TECHNOLOGICAL INSTITUTE TALK
 
 
Guest Speaker
 
Speaker:  Fernando De la Torre, Carnegie Mellon University
Speaker's homepage:  http://www.salleurl.edu/~ftorre/
 
Time:  Wednesday, March 30th @ 1:30pm
Location:  TTI-C Conference Room
 
Title:  "Component Analysis for Multimodal Diaries"
 
Abstract:  
Personalized agents that can monitor and reason about our daily activity
can assist us in many ways (e.g. improving our time management, helping
to have better knowledge of ourselves), and they have a wide range of
applications in other areas such as surveillance and health care
monitoring. In this talk, I will describe preliminary work on contextual
and physical awareness systems that record audio, omnidirectional video,
body sensing data and computer monitor programs in the context of
meeting understanding and office monitoring. I will discuss several
statistical learning algorithms (generative, discriminative and
clustering) able to process and learn from huge amounts of high
dimensional data coming from long-term multimodal sensing.  In
particular, I will describe in a unified framework four component
analysis techniques:

1) Robust parameterized component Analysis (RPCA):
Extension of principal component analysis (PCA) to build a linear model
that is robust to outliers and invariant to geometric transformations.

2) Multimodal Oriented Component Analysis (MODA):
Generalization of linear discriminant analysis (LDA) optimal for
Gaussian multimodal classes with different covariances.

3) Representational Oriented Component Analysis(ROCA):
Extension of OCA to improve classification accuracy when few training
samples are available. (e.g. just 1 training sample). 

4) Discriminative Cluster Analysis(DCA):
Unsupervised low dimensional reduction method that finds a subspace
where clustering with the K-means algorithm performs better.

Applications of these techniques to visual tracking, learning and
recognition, background modeling and temporal segmentation of activities
from multimodal data in the context of meeting understanding and office
monitoring will be discussed.
 
 
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.uchicago.edu/pipermail/colloquium/attachments/20050329/e4019ec5/attachment.htm