Thread Subject: interp3 question?

I have a set of points; say D1 (3x1), Each of these points
is assigned a matrix, F (20 x 3).

I also have another set of points, say D2 (4 x 1), how can
I interpolate I wish to interpolate F over the set D2.

I looked up interp3 and griddatan, but was not able to
accomplish my goal, any tips?

In article <g2i9o9$qsu$1@fred.mathworks.com>,
Hydroman S <amirgsalem@gmail.com> wrote:
>I have a set of points; say D1 (3x1), Each of these points
>is assigned a matrix, F (20 x 3).

>I also have another set of points, say D2 (4 x 1), how can
>I interpolate I wish to interpolate F over the set D2.

Perhaps if you gave an example of your data structures?
Most of what you say has left me uncertain about what you want to
do, but rather than point out the possible interpretations, it
would be easier if you gave an example.
--
  "After all, what problems has intellectualism ever solved?"
                                              -- Robert Gilman

the example:

D1=rand(3,1);
F=rand(20,3);
D2=rand(4,1);

and I wish to interpolate F over the set D2.



I tried:

interp1((meshgrid(D1,ones(1,length(F)))),F,D2), but I get
the error: ??? Error using ==> interp1
The values of X should be distinct.

I also tried interp3 and griddatan to account for the
dimensionality of F, but I also ended up with errors.

In article <g2ibrp$at0$1@fred.mathworks.com>,
Hydroman S <amirgsalem@gmail.com> wrote:
>the example:

>D1=rand(3,1);
>F=rand(20,3);
>D2=rand(4,1);

Your original question said that *each* of the points in D1 was
assigned a 20 x 3 matrix, F, but in this example, there is only
-one- 20 x 3 matrix.

Is each row of D1 associated with a single column of F?
So D1(2) is associated with F(:,2) ? And for any particular D2(K),
you are expecting to produce 20 output values?

Are the rows of F independant of each other, or is there some kind
of cross-correlation between the rows?
--
  "Pray do not take the pains / To set me right. /
   In vain my faults ye quote; / I wrote as others wrote /
   On Sunium's hight." -- Walter Savage Landor

"Hydroman S" <amirgsalem@gmail.com> wrote in message
<g2i9o9$qsu$1@fred.mathworks.com>...
> I have a set of points; say D1 (3x1), Each of these points
> is assigned a matrix, F (20 x 3).
>
> I also have another set of points, say D2 (4 x 1), how can
> I interpolate I wish to interpolate F over the set D2.
>
> I looked up interp3 and griddatan, but was not able to
> accomplish my goal, any tips?
>

You have three points, in a one dimensional
domain space, that map to three function
values. So you use interp1. It so happens that
you have 20 of these relationships to interpolate,
and conveniently for you, interp1 will do them
all in one call. Try it.

You do not have an n-dimensional problem
that needs a high dimensional interpolation.

Use interp1.

John

Thanks Walter and John.

Walter:

> Is each row of D1 associated with a single column of F?

Yes.

> So D1(2) is associated with F(:,2) ?

yes.


And for any particular D2(K),
> you are expecting to produce 20 output values?

Yes.


>
> Are the rows of F independant of each other, or is there
some kind
> of cross-correlation between the rows?

Yes the rows of F are independant of each other.



roberson@ibd.nrc-cnrc.gc.ca (Walter Roberson) wrote in
message <g2imq4$hd$1@canopus.cc.umanitoba.ca>...
> In article <g2ibrp$at0$1@fred.mathworks.com>,
> Hydroman S <amirgsalem@gmail.com> wrote:
> >the example:
>
> >D1=rand(3,1);
> >F=rand(20,3);
> >D2=rand(4,1);
>
> Your original question said that *each* of the points in
D1 was
> assigned a 20 x 3 matrix, F, but in this example, there
is only
> -one- 20 x 3 matrix.
>
> Is each row of D1 associated with a single column of F?
> So D1(2) is associated with F(:,2) ? And for any
particular D2(K),
> you are expecting to produce 20 output values?
>
> Are the rows of F independant of each other, or is there
some kind
> of cross-correlation between the rows?
> --
> "Pray do not take the pains / To set me right. /
> In vain my faults ye quote; / I wrote as others
wrote /
> On Sunium's hight." -- Walter
Savage Landor

"Hydroman S" <amirgsalem@gmail.com> wrote in message
<g2j5kq$4cl$1@fred.mathworks.com>...
> Thanks Walter and John.
>
> Walter:
>
> > Is each row of D1 associated with a single column of F?
>
> Yes.
>
> > So D1(2) is associated with F(:,2) ?
>
> yes.
>
>
> And for any particular D2(K),
> > you are expecting to produce 20 output values?
>
> Yes.
>
>
> >
> > Are the rows of F independant of each other, or is there
> some kind
> > of cross-correlation between the rows?
>
> Yes the rows of F are independant of each other.

Then read my response. Use interp1.

John

John, are you sugessting interpolating over each row of F
in a loop:

for i=1:length (F)
FNEW(i,:)=interp1(D1,F(i,:),D2')
end




"John D'Errico" <woodchips@rochester.rr.com> wrote in
message <g2ir1d$bh9$1@fred.mathworks.com>...
> "Hydroman S" <amirgsalem@gmail.com> wrote in message
> <g2i9o9$qsu$1@fred.mathworks.com>...
> > I have a set of points; say D1 (3x1), Each of these
points
> > is assigned a matrix, F (20 x 3).
> >
> > I also have another set of points, say D2 (4 x 1), how
can
> > I interpolate I wish to interpolate F over the set
D2.
> >
> > I looked up interp3 and griddatan, but was not able to
> > accomplish my goal, any tips?
> >
>
> You have three points, in a one dimensional
> domain space, that map to three function
> values. So you use interp1. It so happens that
> you have 20 of these relationships to interpolate,
> and conveniently for you, interp1 will do them
> all in one call. Try it.
>
> You do not have an n-dimensional problem
> that needs a high dimensional interpolation.
>
> Use interp1.
>
> John

"Hydroman S" <amirgsalem@gmail.com> wrote in message
<g2j6qa$gd2$1@fred.mathworks.com>...
> John, are you sugessting interpolating over each row of F
> in a loop:
>
> for i=1:length (F)
> FNEW(i,:)=interp1(D1,F(i,:),D2')
> end

No. I never said that, although you could
do so. I did state this...

> It so happens that
> you have 20 of these relationships to interpolate,
> and conveniently for you, interp1 will do them
> all in one call. Try it.

Does nobody ever believe me? Why would I lie?

interp1(linspace(0,1,5)',rand(5,6),rand(3,1))
ans =
      0.77177 0.14707 0.76589 0.81177 0.56821 0.21676
       0.8487 0.57487 0.13975 0.57749 0.20276 0.34827
      0.85734 0.51283 0.13816 0.69665 0.14793 0.37811

A one dimensional problem, but here with
6 outputs.

By the way, if you are seriously interpolating
with only three data points, don't expect a
terribly good prediction.

John

this works, thanks John. I'm a believer now :)

I guess what was confusing me is that the length of D had
to be equal to the # of columns in F as you show in your
example:

interp1(linspace(0,1,5)',rand(5,6),rand(3,1))