BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:From calcium imaging to spikes\, using sequential Monte Carlo meth
 ods - Joshua Vogelstein\, Johns Hopkins University
DTSTART:20081001T101500Z
DTEND:20081001T111500Z
UID:TALK13808@talks.cam.ac.uk
CONTACT:Rachel Fogg
DESCRIPTION:Great technological and experimental advances have recently fa
 cilitated the imaging neural activity both in live animals.  We describe a
  sequential Monte Carlo (SMC) expectation maximization algorithm that both
  infers the posterior distributions of the hidden states\, and finds the m
 aximum likelihood estimates of the parameters. Using such an approach enab
 les us to (i) incorporate errorbars on the estimate of the hidden states\,
  (ii) allow for nonlinearities in the observation and transition distribut
 ions\, and (iii) consider Markov priors governing neural activity. This st
 rategy works in real time for each observable neuron. We show how this met
 hod can condition the inferred spike trains on external stimuli\, and achi
 eve superresolution\, i.e.\, infer not just whether a spike occurred withi
 n a stimulus frame\, but when within that frame. Furthermore\, our model h
 as a relatively small number of parameters\, and each of the parameters ma
 y be estimated using standard gradient ascent techniques\, without needing
  "ground truth". We demonstrate the advantage of this approach over the Wi
 ener and a nonnegative deconvolution filter using data sets containing gro
 und truth.\n
LOCATION:LR12\, Engineering\, Department of
END:VEVENT
END:VCALENDAR