Print

Print


Hi Saad,

I was referring to the time before the --rseed option existed and I tried to
do the parallelization below and got 5 identical 5K results.  

Peace,

Matt.

-----Original Message-----
From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On Behalf
Of Saad Jbabdi
Sent: Wednesday, August 12, 2009 3:23 AM
To: [log in to unmask]
Subject: Re: [FSL] waytotal results

Hi Matt,

Yes you give it an integer (default is 12345) - each different integer  
will trigger a different random sequence.
And yes, you can use this option to parallelise your samples. (it is  
working -I just tried it- I don't remember days when it didn't?)

Cheers,
Saad.

On 12 Aug 2009, at 04:47, Matt Glasser wrote:

> Hi Saad,
>
> I am curious about how the --rseed option works.  Do you give it a  
> number to
> seed a random generator, or does it use a random number to seed the  
> random
> generator?  I am wondering if I could use this to parallelize  
> probtrackx on
> a number of samples basis.  For example, if I wanted to run 25000  
> samples
> for each seed, could I break it up into 5 unique --rseed results of  
> 5000
> each and then sum them?  I remember that this did not used to be  
> possible.
>
> Thanks,
>
> Matt.
>
> -----Original Message-----
> From: FSL - FMRIB's Software Library [mailto:[log in to unmask]] On  
> Behalf
> Of Saad Jbabdi
> Sent: Tuesday, August 11, 2009 12:08 PM
> To: [log in to unmask]
> Subject: Re: [FSL] waytotal results
>
> Hi Rachel,
>
> Is this consistent across trials? (i.e. if you re-run probtrackx for
> different --rseed options, do you always get the same pattern?).
>
> Note: it should be the opposite on average! The drop in waytotal
> represents the following probabilities:
>
> 76 = probability of passing through an exclusion mask
> 30474 = probability of NOT passing through an inclusion mask
>
> 40118 = probability of either passing through an exclusion mask or NOT
> passing through an inclusion mask
>
> So the sum of the first two should be larger than the third on average
> because you are counting twice the events where you either pass
> through an exclusion mask or do not pass through an inclusion mask...
>
>
> Cheers,
> Saad.
>
>
>
>
> On 11 Aug 2009, at 17:29, Rachel Hawe wrote:
>
>> Thanks for your response.  What I am more confused about is that
>> adding only
>> an exclusion mask caused a drop of 76 in the waytotal; adding only a
>> waypoint mask caused a drop of 30474; but adding both caused a drop  
>> of
>> 40118.  Why is the drop in waytotal greater when using both an
>> exclusion and
>> waypoint mask compared to the sum of the drops when adding each
>> separately
>> (i.e. 76 + 30474 < 40118)?
>>
>> Thanks,
>> Rachel
>>
>
> --
> Saad Jbabdi
> University of Oxford, FMRIB Centre
>
> JR Hospital, Headington, OX3 9DU, UK
> +44 (0) 1865 222523  (fax 717)
> www.fmrib.ox.ac.uk/~saad
>

--
Saad Jbabdi
University of Oxford, FMRIB Centre

JR Hospital, Headington, OX3 9DU, UK
+44 (0) 1865 222523  (fax 717)
www.fmrib.ox.ac.uk/~saad