Next: Model
Up: Introduction
Previous: Introduction
We start in section 2 by describing how the GLM in a
fully Bayesian framework allows us to introduce soft constraints
on linear combinations within the GLM. We then discuss how we
choose a sensible basis set, and determine the required basis set
constraints. In section 3 we describe how we
perform approximate Variational Bayesian inference on the model.
In section 4 we use the model on artificial null data
to demonstrate the effect of the HRF constraints. Then in
section 5 we describe how we can use spatial mixture
modelling to produce probabilities of activation which takes
advantage of the extra sensitivity produced from the HRF
constraints. Finally, in section 6 we demonstrate
this increased sensitivity on an audio-visual dataset.