Hi
The ARD weight is a fudge factor. The prior (energy) in bedpostx
write: prior_en = -log(prior) = fudge_factor * log(f)
So increasing the weight (fudge factor) decreases the energy, allowing
f to explore higher values. I.e. more ard fudge factor means LESS
constraints for f to be small.
We set this weight to be equal to 1 by default.
Cheers,
Saad.
On 22 May 2009, at 08:34, SUBSCRIBE FSL Anonymous wrote:
> Hello,
>
> I have a question about "ARD in bedpostx".
>
> Bedpostx use ARD on the volume fractions of all but the first fiber.
> I 've read bedpostx program.
> The posterior energy is computed as
> "posterior_en = prior_en( f_prior,etc) + likelihood_en(error beween
> m_data and
> pred)".
> The prior of volume fractions is computed as
> "f_prior = ARD(default=1.0) * log(f)"
> and f_prior is a value of the minus.
>
> I assume that if ARD(weight) is a small value, posterior_en is
> large. As a
> result, the influence of f2,f3.. increases.
> Does ARD work anything else ?
>
> I appreciate your king response very much.
>
--
Saad Jbabdi
Oxford University FMRIB Centre
JR Hospital, Headington, OX3 9DU, UK
+44 (0) 1865 222545 (fax 717)
www.fmrib.ox.ac.uk/~saad
|