BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A tale of P-matrices and TripleSpinners - the unreasonable effecti
 veness of structured models in nonlinear embeddings - Krzysztof Choromansk
 i\, Google NY
DTSTART:20161124T110000Z
DTEND:20161124T120000Z
UID:TALK67544@talks.cam.ac.uk
CONTACT:Adrian Weller
DESCRIPTION:In this talk we will present new generic paradigms for conduct
 ing machine learning computations with structured linear projections. Thes
 e paradigms can be applied to speed up several machine learning algorithms
  based on random linear projections such as: cross-polytope LSH techniques
 \, kernel approximations via random feature maps\, quantization methods us
 ing random projection trees\, several variants of Johnson-Lindenstrauss tr
 ansforms\, and many more.\n \nTheir adaptive versions can be applied in ne
 ural network architectures to construct much more compact and faster yet s
 till good quality neural network models. As a byproduct\, we give the firs
 t theoretical guarantees regarding the fastest known cross-polytope LSH me
 thods based on the Walsh-Hadamard Transform.\n\nThe proposed structured fa
 milies of P-model and TripleSpin matrices cover as special cases all struc
 tured matrices used so far in this context\, but also include new structur
 ed constructions not considered previously.\n\n\nBio: \nKrzysztof Choroman
 ski is a member of the Google Brain Robotics Team in New York. He works on
  several problems regarding robotics and machine learning such as: reinfor
 cement learning\, control theory\, state estimation\, predictive state rep
 resentation and Hilbert embeddings of dynamical systems. His research inte
 rests include also\nneural networks and the theory of structured nonlinear
  embeddings. The latter are used in several machine learning applications 
 such as: compact neural networks with fast inference\, kernel approximatio
 n techniques via random feature maps\, and the fastest known cross-polytop
 e LSH algorithms. Krzysztof plays piano and is an avid salsa dancer.
LOCATION:CBL Room BE-438
END:VEVENT
END:VCALENDAR
