BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Gaussian Process Latent Variable Models - Marc Deisenroth (Univers
 ity of Cambridge)
DTSTART:20081113T140000Z
DTEND:20081113T153000Z
UID:TALK13499@talks.cam.ac.uk
CONTACT:Shakir Mohamed
DESCRIPTION:Summarizing a high dimensional data set with a low dimensional
  embedding \nis a standard approach for exploring its structure. In this p
 aper we \nprovide an overview of some existing techniques for discovering 
 such \nembeddings. We then introduce a novel probabilistic interpretation 
 of \nprincipal component analysis (PCA) that we term dual probabilistic PC
 A \n(DPPCA). The DPPCA model has the additional advantage that the linear 
 \nmappings from the embedded space can easily be non-linearized through \n
 Gaussian processes. We refer to this model as a Gaussian process latent \n
 variable model (GP-LVM). Through analysis of the GP-LVM objective \nfuncti
 on\, we relate the model to popular spectral techniques such as \nkernel P
 CA and multidimensional scaling. We then review a practical \nalgorithm fo
 r GP-LVMs in the context of large data sets and develop it \nto also handl
 e discrete valued data and missing attributes. We \ndemonstrate the model 
 on a range of real-world and artificially \ngenerated data sets.\n\nparts 
 from these papers will be discussed:\n\nhttp://www.jmlr.org/papers/volume6
 /lawrence05a/lawrence05a.pdf\n\n"ftp://ftp.dcs.shef.ac.uk/home/neil/gplvmT
 utorial.pdf":URL
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR