Hello
Søren,Perhaps there was a miscommunication. Yes, time is sampled at uniform
intervals, like you pointed out. What I’m talking about is in parameter study.
Let me give an example with pseudo-code.Main = { … Study = { …
AnyVar R = 1.0; }; // end study}; // end mainparameterStudy = {…AnyDesVar P =
{ Val = Main.Study.R; Min = 0.9Val; Max = 1.1Val;}nSteps =
{3};Every time I run the parameter study, the analysis would be run 3 times, for
R values in {0.9, 1.0, 1.1}. This is because the closed interval [0.91.0,
1.11.0] is sampled at equal intervals of 0.1. These are the same values every
time I run.The set {0.95, 1.0, 1.05} is also uniformly distributed. It just
doesn’t include the extreme points. Put another way, sampling at equal intervals
leads to a uniform distribution, but not all uniform distributions have samples
at equal intervals. I’ve tried to give a more set theoretic explanation further
below, if it would help.The point I’m trying to get across is that every time I
run a parameter study simulation, I should be able to obtain different sequences
for DesVar that are not necessarily sampled at the same equal intervals of
(Max-Min) / (nSteps-1).Sincerely,-Venn-----------------------------------In
terms of set the theory,Let U be the set of all uniformly distributed 3-tuples
from the closed interval [0.9, 1.1], i.e, U = { (x1,x2,x3) | 0< = x1, x2, x3
<= 1.1 and (x1, x2, x3) are uniformly distributed } Note that U is an
uncountably infinite set.Let e be the 3-tuple that is equally distributed from
the same closed interval , as is the case with Anybody, i.e., e = (0.9, 1.0,
1.1);Clearly, {e } is only a subset of U. In fact e is but a trivial member of
U .Additionally, we could see that e belongs to E where E is the (uncountably
infinite) set of all equally distributed 3-tuples from the same closed interval
[0.9, 1.1]. E is also a subset of U. Example: (0.9, 1.01, 1.1) belongs to U
but not to E, while (0.93, 0.96, 0.99) and (0.9, 0.95, 1.0) belong to E and
in turn to U. To: anyscript@yahoogroups.comFrom: support@anybodytech.comDate:
Thu, 22 Feb 2007 09:13:04 +0000Subject: [anyscript] Re: Parameter Study
Hi Venn
I am not sure i understand you compleately, i will try to explain how
it works.
The property nStep simply tell how many step to use
tStart is the time value when the analysis starts
tEnd is the time value when it ends.
Example1:
nStep=3;
tEnd=1;
tStart=0;
The analysis will calulate for the following time values: t={0,0.5,1}
Example2:
nStep=5;
tEnd=3;
tStart=1;
The analysis will calulate for the following time values: t=
{1,1.5,2,2.5,3}
So there is not any propability in play here, it is a uniform
distribution of time values between tStart and tEnd divided into
nSteps
Best regards
Søren
– In anyscript@yahoogroups.com, Vengateswaran Ravichandran
<vennjr@…> wrote:
>
>
> Søren,Using a dR term is a much more elegant way of doing what I
was trying to do. Thanks for your help. I’ll implement it.Now, about
Question 1b.You are right that the default values generated by
Anybody would fit a uniform distribution. But it is not the only
uniform distribution I could generate.If I obtain 11 numbers randomly
generated in the interval [0,10], such that the resulting numbers are
uniformly distributed, I agree that the numbers {0, 1, 2, …, 10}
would fit the bill. But uniform distribution means that any number in
the interval [0,11] is equally likely. In other words the probability
for any number (within [0,11]) to be chosen is equal to the
probability of any other number being chosen. However, when I choose
{0, 1,…, 10} every time I try to chose a set of 11 numbers, the
probability is not the same. Simply put, I’m trying to differentiate
between when the intervals between numbers chosen are equal and when
the probability of any number being chosen are equal. I don’t know if
I made sense… if I did not, please let me know and I’ll try to
explain better.Sincerely,-VennTo: anyscript@…: support@…: Tue, 20
Feb 2007 09:17:46 +0000Subject: [anyscript] Re: Parameter Study
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> Hi Venn
>
> I have taken a look at your file, you are right what you are trying
> to do there is not possible. The reason is that you are trying to
> assign Radius using both a design variable and an expression which
is
> not independent, this can not be done, the system would not known
> which of them to use. A possible hack could be if you use the
normal
> espression for Radius and then in that expresion add the value of
> your design variable. Then you would have the normal radius and the
> parameter study could do changes to this value
>
> it would look something like this:
>
> dR=0; //from your paramter study make a lnk to this varaible
>
> Radius = sqrt((.node1.sRel[0]-…P1[0])^2 + (.node1.sRel[1]-…P1[1])
> ^2 + (.node1.sRel[2]-…P1[2])^2)+ dR;
>
> In this way you can change the radius using your design variable
and
> if it is zero you have not made any change.
>
> About question 1a i have made a small example which should
illustrate
> this, it has been uploaded as “ParamterStudyWithSelectedValues.any”
>
> About question 1b this what AnyBody will do by default, if you
> specify a nStep={10} and your design variable has a min =0 and max=
1
> it will distribute the 10 steps uniformly between 0 and 1.
>
> Best regards
> Søren
>
>
> — In anyscript@yahoogroups.com, Vengateswaran Ravichandran
> <vennjr@> wrote:
> >
> > Thanks
> > Søren,I have done the same thing as you suggested (in
> SufParamStudy.any). However, this means that I’ve to modify some of
> the files in the repository, which I was hoping to avoid. I’ve
> posted a file (SurfParamStudy-Venn.any) that shows an example of
what
> I’m trying to do to the group that better illustrates my
> predicament.For question 1a, can you give me an example. I’m not
sure
> how I’ll be able to use t and Data with design variables.For
question
> 1b, what I meant was this. Instead of me giving the exact data set,
I
> could specify a range [0,10] and ask for say 20 samples. It would
be
> nice if Anybody can pull out these 20 samples from an internally
> generated uniform distribution.Thanks again for the speedy
> response,To: anyscript@: support@: Mon, 19 Feb 2007 10:57:55
> +0000Subject: [anyscript] Re: Parameter Study
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> > Hi Venn
> >
> > Thanks for the your suggestions and kind words.
> >
> > It is possible to use the dimensions of a paramsurf as design
> > variable, i have made very simple example illustrating this i
will
> > upload to the file section as “SurfParamStudy.any”, shortly. It
> > has one design variable which change the radius of a cylinder on
> > which a line is wrapping. There is one thing however, the surface
> do
> > not visually change size in the modelview, but if you look at the
> > muscle you can see, it change its size.
> >
> > About your question 1a:
> > It is possible to do a parameter study where the grid is unevenly
> > distributed. This can be done by using a piecewise linear
> > interpoaltion function and use this as a “translation table”. So
> you
> > let your design variable go from for example 0 to 10 and use
> > nstep=10, then you create the interpolation function with t=
> > {0,1,2…9} and Data={Selected values…!}.
> >
> > Question 1b, i am not sure i understand what you mean by uniform
> > distribution in this relation.
> >
> > Question 2: we will consider this for the next release.
> >
> > Qestion 3: This is not possible today because the FileName
property
> > of the Outputfile is a constant, which can not be changed after
> > loadtime, we will also consider to open up for this is the next
> > releases.
> >
> > Again, thanks for your suggestions!
> >
> > Best regards
> > Søren
> >
> > — In anyscript@yahoogroups.com, “Vengateswaran J. Ravichandran”
> > <vennjr@> wrote:
> > >
> > > Hello,
> > >
> > > First of all, I’m glad you included Parameter Study in AnyBody
> 2.0.
> > > Because it seems that I cannot change the parameters otherwise,
> > without
> > > creating a new model by writing a file with all the parameters.
> > However,
> > > it seems that I cannot vary parameters that are calculated
rather
> > than
> > > assigned. For example, I’m not able to set the radius of
wrapping
> > > surfaces as a AnyDesVar.
> > > I was trying to set
> > >
> >
>
ShoulderArm.Mus.Pectoralis_minor2_cyl.Cylinder.Cyl.node1.CylSurf.Radiu
> > s,
> > > among others. But I get an error. Is there an alternate way to
do
> > this?
> > > I really appreciate your input.
> > >
> > >
> > >
> > > Here are a few of my thoughts on what more could be done to
> improve
> > > Parameter Study in the next version:
> > > 1a. The rectangular grid function is not very helpful when
> > simulating a
> > > lot of parameters (204, in my case). Perhaps it would help to
> > explain
> > > what I’m doing. Clearly, I do not want to do an exhaustive
> search,
> > since
> > > it will take eons to complete. Perhaps I could specify
specific
> > values
> > > instead, say as a vector?
> > > 1b. Even better would be when I AnyBody can pick a uniform
> sampling
> > of
> > > the space specified. Naturally, this would be an option, rather
> > than the
> > > default. (I’m using MATLAB as a workaround to do just this).
> > >
> > > 2. In what I’m doing, there is also no design measure to
maximize
> or
> > > minimize. So, as of now, I’ve created a dummy DesMeasure that
is
> > set to
> > > zero, since it appears mandatory to have a DesMeasure. I don’t
> > know if
> > > DesMeasure needs to be mandatory. Not an important issue, but
> > thought
> > > I’d mention.
> > >
> > > 3. I think it would be helpful if there is a way to save the
> > results of
> > > ParamStudy into a different file after each run. Let me try to
> > explain
> > > with an example.
> > > Let us say ParamStudy runs KinematicAnalysis that outputs a
> > particular
> > > variable to an OutFile. I don’t want to overwrite OutFile each
> time,
> > > instead it would be great if a new OutFile, say with a suffix
> > indicating
> > > the ParamStudy run number, can be created.
> > >
> > > My thanks to all the folks at the support group. My particular
> > thanks to
> > > Søren for all his help.
> > > Your fast turn around time for queries still amazes me. Thank
you
> > for
> > > your patience while I still try to putz around.
> > >
> > > -Venn
> > >
> > >
> > >
> > > [Non-text portions of this message have been removed]
> > >
> >
> >
> >
> > __________________________________________________________
> > Get the most out of tax shelters and other breaks.
> >
>
http://articles.moneycentral.msn.com/Taxes/TaxShelters/TaxShelters.asp
> x?icid=WLtextFebtagline
> >
> > [Non-text portions of this message have been removed]
> >
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
>
> __________________________________________________________
> Invite your mail contacts to join your friends list with Windows
Live Spaces. It’s easy!
> http://spaces.live.com/spacesapi.aspx?
wx_action=create&wx_url=/friends.aspx&mkt=en-us
>
> [Non-text portions of this message have been removed]
>
Discover the new Windows Vista
http://search.msn.com/results.aspx?q=windows+vista&mkt=en-US&form=QBRE
[Non-text portions of this message have been removed]