BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:A talk of two distinct parts: 1) a parametric empirical Bayesian a
 pproach to integration of fMRI\, EEG and MEG data\, 2) prediction error in
  episodic memory encoding - Rik Henson ( MRC Cognition and Brain Sciences 
 Unit)
DTSTART:20131028T110000Z
DTEND:20131028T120000Z
UID:TALK48532@talks.cam.ac.uk
CONTACT:Prof Máté Lengyel
DESCRIPTION:In an attempt to find at least one patch of common ground\, I 
 will start by talking about a recent methodological approach to multimodal
  integration of human neuroimaging data (specifically EEG\, MEG and fMRI) 
 that uses a parametric empirical Bayesian framework (recently reviewed in 
 Henson et al\, 2011\, Frontiers in Human Neuroscience). Using a linear\, h
 ierarchical model of the M/EEG inverse problem under Gaussian assumptions:
  1) MEG and EEG are fused symmetrically\, with separate noise regularisati
 ons (hyperparameters) estimated from the data\, and 2) M/EEG and fMRI are 
 integrated asymmterically\, with separate fMRI clusters forming separate s
 patial priors. Then in the fifth sixth of the talk\, I will switch topics 
 completely to talk about recent behavioural experiments that test the role
  of prediction error in one-shot associative encoding\, as a model of huma
 n episodic memory\, based on our PIMMS (Predictive Interactive Multiple Me
 mory Signals) framework for understanding the neuroscience of memory and p
 erception (Henson & Gagnepain\, Hippocampus\, 2010). I will then ask for y
 our help in understanding why we cannot fit these data with various simple
  Hebbian learning rules.
LOCATION:Engineering Department\, CBL Room 438
END:VEVENT
END:VCALENDAR