BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Computational Neuroscience Journal Club - Yan Wu (University of Ca
 mbridge)
DTSTART:20141021T150000Z
DTEND:20141021T160000Z
UID:TALK55626@talks.cam.ac.uk
CONTACT:Guillaume Hennequin
DESCRIPTION:Yan Wu will cover: \n\nBayesian Active Learning of Neural Firi
 ng Rate Maps with Transformed Gaussian Process Priors\n\nM Park\, J P Well
 er\, G D Horwitz and J Pillow\; Neural Computation\, 2014\n\nhttp://www.mi
 tpressjournals.org/doi/abs/10.1162/NECO_a_00615#.VEC7FHXXphF\n\nA firing r
 ate map\, also known as a tuning curve\, describes the nonlinear relations
 hip between a neuron's spike rate and a low-dimensional stimulus (e.g.\, o
 rientation\, head direction\, contrast\, color). Here we investigate Bayes
 ian active learning methods for estimating firing rate maps in closed-loop
  neurophysiology experiments. These methods can accelerate the characteriz
 ation of such maps through the intelligent\, adaptive selection of stimuli
 . Specifically\, we explore the manner in which the prior and utility func
 tion used in Bayesian active learning affect stimulus selection and perfor
 mance. Our approach relies on a flexible model that involves a nonlinearly
  transformed gaussian process (GP) prior over maps and conditionally Poiss
 on spiking. We show that infomax learning\, which selects stimuli to maxim
 ize the information gain about the firing rate map\, exhibits strong depen
 dence on the seemingly innocuous choice of nonlinear transformation functi
 on. We derive an alternate utility function that selects stimuli to minimi
 ze the average posterior variance of the firing rate map and analyze the s
 urprising relationship between prior parameterization\, stimulus selection
 \, and active learning performance in GP-Poisson models. We apply these me
 thods to color tuning measurements of neurons in macaque primary visual co
 rtex.
LOCATION:Cambridge University Engineering Department\, CBL\, BE-438 (http:
 //learning.eng.cam.ac.uk/Public/Directions)
END:VEVENT
END:VCALENDAR