BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//talks.cam.ac.uk//v3//EN
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:DAYLIGHT
TZOFFSETFROM:+0000
TZOFFSETTO:+0100
TZNAME:BST
DTSTART:19700329T010000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=-1SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:+0100
TZOFFSETTO:+0000
TZNAME:GMT
DTSTART:19701025T020000
RRULE:FREQ=YEARLY;BYMONTH=10;BYDAY=-1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CATEGORIES:Signal Processing and Communications Lab Seminars
SUMMARY:Dual-to-kernel learning with ideals - Dr. Franz Ki
raly\, University College London
DTSTART;TZID=Europe/London:20140313T140000
DTEND;TZID=Europe/London:20140313T150000
UID:TALK50833AThttp://talks.cam.ac.uk
URL:http://talks.cam.ac.uk/talk/index/50833
DESCRIPTION:We propose a theory unifying kernel learning and s
ymbolic algebraic methods. Kernel methods are a ve
ry popular class of algorithms employing kernel fu
nctions which allow to capture properties of the d
ata in a very efficient way\, representing them im
plicitly in the so-called feature space\, the most
prominent example being the kernel support vector
machine. The main advantage of kernels is also th
eir main downside: since the representation is imp
licit it has remained an open question what exactl
y the structures and features are which make the a
lgorithms work.\n\nSymbolic algebraic methods\, on
the other hand\, are by construction structural a
nd deal with the manipulation of explicit equation
s. So far\, their theoretical complexity and intra
ctable computational cost\, such as for GrÃ¶bner ba
sis computations\, has prevented broad application
to real-world learning and data analysis. \n\nWe
show that kernel learning and symbolic algebra are
inherently dual to each other\, and we use this d
uality to combine the structure-awareness of algeb
raic methods with the efficiency and generality of
kernels. The main idea lies in relating polynomia
l rings to feature space\, and ideals to manifolds
\, then exploiting this generative-discriminative
duality on kernel matrices. We illustrate this by
proposing two algorithms\, IPCA and AVICA\, for si
multaneous manifold and feature learning.
LOCATION:LR5\, Cambridge University Engineering Department
CONTACT:Dr Ramji Venkataramanan
END:VEVENT
END:VCALENDAR