-------- Original Message --------
Subject: Re: Modularity: Evaluate hypothesis in MorphoJ
Date: Wed, 8 Feb 2012 03:54:42 -0500
From: Chris Klingenberg <[email protected]>
Reply-To: [email protected]
Organization: University of Manchester
To: [email protected]

Dear Helmi

Hmm, I can only guess what makes your computer grind away like this, but
here is what I suspect.

You start with 49 landmarks and your hypothesis divides them into two
modules of 18 and 49. There are a lot of ways of doing that (the
combinatorial calculator on the web that comes up first on Google says
that there are something like 1.155E13 possiblities: 11.55 trillion!).

So if you choose to do the complete enumeration of all possible
partition, your computer will be busy for a long time! OK, the
restriction to contiguous partitions reduces the possible number of
allowed partitions, but that comes with some extra computational effort.

The message is: use a smaller number of partitions. For example, start
with the default of 10,000 and see how long it takes. Then maybe try
with 100,000 or a million (which will take about 10 or 100 times as
long) -- almost surely, the distributions of RV coefficients will be
very similar.

I hope this is helpful.

Best wishes,
Chris



On 2/7/2012 9:57 PM, morphmet wrote:


-------- Original Message --------
Subject: Modularity: Evaluate hypothesis in MorphoJ
Date: Fri, 3 Feb 2012 05:28:08 -0500
From: hmi hmi <[email protected]>
To: morphomet morphometrics.org <[email protected]>



Dear Morphometricans,

I am trying the modularity: evaluate hypothesis function following the
user guide in MorphoJ (1.03d and 1.04a) on two separate Windows 7-64bit
machines (core-i5 8gb ram and core2 quad 8gb ram-both with 4 CPU cores)
for both 2D and 3D data. Both have the 32-bit java console installed.
The calculations work well for 2D data (25 landmarks) with few seconds
lag. The characteristics of the 3D data are:

49 landmarks (80 specimens) with two partitions, where 18 landmarks are
placed in the second partition. The contiguous partitions only box
ticked and full enumeration of partitions selected. The 49 landmarks
originated from a 200 landmark (80 specimens) morphoJ file.

After selecting this operation, I notice the total CPU usage of the
javaw.exe*32.exe in the task manager goes up to 25% and stays at 25% for
one modularity operation. So if I made another modularity test, the CPU
usage will be 50%. I do not see anything come out in the "Results" and
"Graphics" tab of MorphoJ. When I move the mouse cursor on MorphoJ, it
does show MorphoJ busy calculating the data. I left the computer on
overnight, I see the CPU usage is still 25% with MorphoJ busy with the
calculations. I've tried different analyis in MorphoJ while waiting for
the modularity test and MorphoJ can work on other data and the output is
generated after few seconds lag which shows that the software is not
frozen. However, no analysis done after the modularity test could be
save as the modularity test itself is not completed. I've also tried
increasing CPU usage process priority to high and realtime but that did
not improve the processing time. There does not seem to be any
difference in the processing time regardless of type of CPU usage. The
ram usage is not affected and stays about 142 mb.

Is there something wrong with my data? Does anyone know how to make the
calculations faster?

Grateful for the help.

Regards,
Helmi Pritam
University of Dundee


--
***************************************************************
Christian Peter Klingenberg
Faculty of Life Sciences
The University of Manchester
Michael Smith Building
Oxford Road
Manchester M13 9PT
United Kingdom

Telephone: +44 161 275 3899
Fax: +44 161 275 5082
E-mail: [email protected]
Web: http://www.flywings.org.uk
Skype: chris_klingenberg
***************************************************************


Reply via email to