Hi
No, the current release version does not run multiple runs. While your list of references indeed discuss benefits of combining multiple runs the field in general has forgotten relevant papers by Rao which show that the ICA solution is unique (see our 2004 paper for a discussion) Probably the biggest culprit for between-run variability in melodic is the convergence criterion - you can reduce between-run variability by reducing the minimum error change to something really small (using the --eps command line option).
hth
Christian
On 16 Mar 2011, at 19:08, Lesley Fidler wrote:
> Hi FSL experts,
>
> I am having a hard time getting my question right. I'll try again. Sorry if I seem a bit dim here but this is a new direction for me.
>
> In several articles the problem of ICAs stochastic approach to detecting the minima has been discussed. Here are a couple (Himberg, Hyvarinen & Esposito, 2004 and Schmithorst, Holland & Plante, 2006). All of these articles discuss algorithms to establish components that have a level of stability.
>
> My question is does MELODC take this into account? If so could you point me to the report? Does MELODIC perform several iterations? If so how is the number controlled?
>
> Hope you can help,
> Thank you very much
> Lesley Fidler
|