Turns out it is easy to create a transfer function using the other
notation, just use the conjugate.

--- In jsci@yahoogroups.com, "alkottke" <alkottke@...> wrote:
>
> Hello,
>
> So I found out the issue.  Matlab uses -2pi in the transform and 2pi
> in the inverse, whereas JSci uses 2pi in the transform and -2pi in the
> inverse transform. If I recall correctly, the difference between +/-
> 2pi is a matter of convention, but I don't know how to create a
> transfer function in the frequency domain when the -2pi convention is
> applied.  Mathworld also uses the -2pi transform, 2pi inverse
> convention: http://mathworld.wolfram.com/FourierTransform.html
>
> --- In jsci@yahoogroups.com, Mark Hale <mj_hale@> wrote:
> >
> > Hi,
> >
> >  As the data is increasing positive, increasing negative then you
> don't need the first initial sort, ie inverseTransform(data), if you
> want the output in sequential time order, then you need the last sort,
> ie sort(inverseTransform(data)).
> >
> >  If you still have problems, send me some test data and a screen
> shot of the results in matlab.
> >
> >  Regards,
> >
> >  Mark
> >
> >
> > ----- Original Message ----
> > From: alkottke <alkottke@>
> > To: jsci@yahoogroups.com
> > Sent: Thursday, 24 August, 2006 8:22:58 PM
> > Subject: [jsci] inverse Fourier Transform
> >
> > I had some problems getting the inverse Fourier Transform to work.  I
> > have a Fourier Amplitude Spectrum (FAS) that with increasing positive
> > frequencies, the Nyquist frequency, and the increasing negative
> > frequencies.  I tried using sort(inverseTransform(sort(...))) to
> > compute the time history, but the results don't make any sense.  I
> > also tried with and without the both of the sort() commands.  Which is
> > appropriate?
> >
> > Right now I decided to use another FFT library because the results
> > that it produces match the results I compute using matlab.  However, I
> > would like to use JSci to simplify the libraries.
> >
> > Can anyone explain to me how to properly use inverseTransform()?
> >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
>

Hello,

So I found out the issue.  Matlab uses -2pi in the transform and 2pi
in the inverse, whereas JSci uses 2pi in the transform and -2pi in the
inverse transform. If I recall correctly, the difference between +/-
2pi is a matter of convention, but I don't know how to create a
transfer function in the frequency domain when the -2pi convention is
applied.  Mathworld also uses the -2pi transform, 2pi inverse
convention: http://mathworld.wolfram.com/FourierTransform.html

--- In jsci@yahoogroups.com, Mark Hale <mj_hale@...> wrote:
>
> Hi,
>
>  As the data is increasing positive, increasing negative then you
don't need the first initial sort, ie inverseTransform(data), if you
want the output in sequential time order, then you need the last sort,
ie sort(inverseTransform(data)).
>
>  If you still have problems, send me some test data and a screen
shot of the results in matlab.
>
>  Regards,
>
>  Mark
>
>
> ----- Original Message ----
> From: alkottke <alkottke@...>
> To: jsci@yahoogroups.com
> Sent: Thursday, 24 August, 2006 8:22:58 PM
> Subject: [jsci] inverse Fourier Transform
>
> I had some problems getting the inverse Fourier Transform to work.  I
> have a Fourier Amplitude Spectrum (FAS) that with increasing positive
> frequencies, the Nyquist frequency, and the increasing negative
> frequencies.  I tried using sort(inverseTransform(sort(...))) to
> compute the time history, but the results don't make any sense.  I
> also tried with and without the both of the sort() commands.  Which is
> appropriate?
>
> Right now I decided to use another FFT library because the results
> that it produces match the results I compute using matlab.  However, I
> would like to use JSci to simplify the libraries.
>
> Can anyone explain to me how to properly use inverseTransform()?
>
>
>
>
>
>
> Yahoo! Groups Links
>

Hi,

  As the data is increasing positive, increasing negative then you don't need the
first initial sort, ie inverseTransform(data), if you want the output in
sequential time order, then you need the last sort, ie
sort(inverseTransform(data)).

  If you still have problems, send me some test data and a screen shot of the
results in matlab.

  Regards,

  Mark


----- Original Message ----
From: alkottke <alkottke@...>
To: jsci@yahoogroups.com
Sent: Thursday, 24 August, 2006 8:22:58 PM
Subject: [jsci] inverse Fourier Transform

I had some problems getting the inverse Fourier Transform to work.  I
have a Fourier Amplitude Spectrum (FAS) that with increasing positive
frequencies, the Nyquist frequency, and the increasing negative
frequencies.  I tried using sort(inverseTransform(sort(...))) to
compute the time history, but the results don't make any sense.  I
also tried with and without the both of the sort() commands.  Which is
appropriate?

Right now I decided to use another FFT library because the results
that it produces match the results I compute using matlab.  However, I
would like to use JSci to simplify the libraries.

Can anyone explain to me how to properly use inverseTransform()?






Yahoo! Groups Links

I had some problems getting the inverse Fourier Transform to work.  I
have a Fourier Amplitude Spectrum (FAS) that with increasing positive
frequencies, the Nyquist frequency, and the increasing negative
frequencies.  I tried using sort(inverseTransform(sort(...))) to
compute the time history, but the results don't make any sense.  I
also tried with and without the both of the sort() commands.  Which is
appropriate?

Right now I decided to use another FFT library because the results
that it produces match the results I compute using matlab.  However, I
would like to use JSci to simplify the libraries.

Can anyone explain to me how to properly use inverseTransform()?

Hi Mark,

I found it.

Thanks,
Steve

--- Mark Hale <mj_hale@...> wrote:

> Hi Steve,
>
>  Yes, it does, see
> JSci.maths.LinearMath.leastSquaresFit or
> .linearRegression.
>
>  Regards,
>  Mark
>
>
> ----- Original Message ----
> From: ssinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Wednesday, 19 July, 2006 7:04:14 PM
> Subject: [jsci] Linear Regression
>
> Does JSci have any any kind of linear
> regression-solving capabilities?
> I searched through, but couldn't find anything.
> It's no big deal if
> it doesn't, as I can write them myself or use
> another math library.
> But since most of the math functions I use come from
> JSci, it's nice
> to be consistent and draw as much as I can from it.
>
> Thanks,
> Steve Sinai
>
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Hi Steve,

  Yes, it does, see JSci.maths.LinearMath.leastSquaresFit or .linearRegression.

  Regards,
  Mark


----- Original Message ----
From: ssinai <ssinai@...>
To: jsci@yahoogroups.com
Sent: Wednesday, 19 July, 2006 7:04:14 PM
Subject: [jsci] Linear Regression

Does JSci have any any kind of linear regression-solving capabilities?
I searched through, but couldn't find anything.  It's no big deal if
it doesn't, as I can write them myself or use another math library.
But since most of the math functions I use come from JSci, it's nice
to be consistent and draw as much as I can from it.

Thanks,
Steve Sinai








Yahoo! Groups Links

Does JSci have any any kind of linear regression-solving capabilities?
I searched through, but couldn't find anything.  It's no big deal if
it doesn't, as I can write them myself or use another math library.
But since most of the math functions I use come from JSci, it's nice
to be consistent and draw as much as I can from it.

Thanks,
Steve Sinai

Hi Mark,

Seems to work!  Thanks.

Steve Sinai

--- Mark Hale <mj_hale@...> wrote:

> Added v0.944.
>
> ----- Original Message ----
> From: Steve Sinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Wednesday, 31 May, 2006 12:10:18 AM
> Subject: Re: [jsci] luDecompose problem in
> DoubleSquareMatrix
>
> Hi Mark,
>
> If it isn't too much trouble, adding the textbook
> version of LU decomposition to JSci would be
> helpful.
> Part of the reason for my editor is as a teaching
> aid,
> so even if the "classic LU decomposition" (as my
> book
> describes it) isn't used much outside of the
> classroom, in this case the classroom version would
> be
> nice.
>
> I guess the phrase "classic LU decomposition" is a
> polite way of saying that it's not too useful for
> modern, real-world applications.
>
> Thanks,
> Steve Sinai
>
> --- Mark Hale <mj_hale@...> wrote:
>
> > Yes, it does.  The LU decomposition is often an
> > intermediary step, so the fact that it is
> permutated
> > (scrambled) usually doesn't matter because you
> > either unscramble at a later step (e.g. matrix
> > inversion), or the final calculation isn't
> affected
> > (so much) by permutations (e.g. matrix
> determinant).
> >  If you are calculating the LU just for its own
> sake
> > (e.g. textbook) then you need an LU method without
> > pivoting...which I can provide if desired.
> >
> > ----- Original Message ----
> > From: Steve Sinai <ssinai@...>
> > To: jsci@yahoogroups.com
> > Sent: Tuesday, 30 May, 2006 10:01:28 AM
> > Subject: Re: [jsci] luDecompose problem in
> > DoubleSquareMatrix
> >
> > Hi Mark,
> >
> > I'm always impressed with your quick responses.
> >
> > Unfortunately, I'm one of those people who hit the
> > math wall at long division, and have no idea what
> > matrices have to do with permutations. I see it's
> > mentioned later in the linear algebra book I'm
> > slogging through, and I'll eventually get to it.
> No
> > need for you to try explaining it to me.  It would
> > be
> > futile.
> >
> > Anyway, I've never actually used an LU
> decomposition
> > for anything.  I'm just trying to write this weird
> > text-based math editor (like MATLAB or MAPLE)
> using
> > JSci, and I'm reading through a linear algebra
> book
> > and testing out the editor using the examples in
> the
> > book.  Does the JSci implementation of LU
> > Decomposition reflect how people in the real world
> > use
> > it? If so, I'll just leave it as is.
> >
> > Thanks,
> > Steve Sinai
> >
> > --- Mark Hale <mj_hale@...> wrote:
> >
> > > Hi,
> > >
> > >  luDecompose performs the LU decomposition up to
> a
> > > permutation given by the pivot array.  That is,
> > the
> > > original matrix M is permutated to a matrix M'
> and
> > > then the LU decomposition is performed on M'.
> M'
> > is
> > > chosen to improve the numerical stability. You
> can
> > > use the pivot array to transform back the output
> > but
> > > ofcourse it won't be lower triangular any
> more...
> > >
> > >  If you disable the pivot code in the source,
> you
> > > should get what you want (but more prone to
> > > numerical instabilities).
> > >
> > >  Regards,
> > >
> > >  Mark
> > >
> > > ----- Original Message ----
> > > From: ssinai <ssinai@...>
> > > To: jsci@yahoogroups.com
> > > Sent: Monday, 29 May, 2006 10:10:05 PM
> > > Subject: [jsci] luDecompose problem in
> > > DoubleSquareMatrix
> > >
> > > Hi,
> > >
> > > I'm trying to use the luDecompose method of
> > > DoubleSquareMatrix, and am
> > > not getting the correct answer to my test
> > problems.
> > > I'm hardly a
> > > master of LU Decomposition, and am using some
> > > example problems from a
> > > linear algebra text.  Anyway, the initial matrix
> > > from the text is
> > >  1  2 -3
> > > -3 -4 13
> > >  2  1 -5
> > > and the answer the text gives is
> > > L= 1    0  0
> > >   -3    1  0
> > >    2 -3/2  1
> > >
> > > U= 1 2 -3
> > >    0 2  4
> > >    0 0  7
> > >
> > > Some test code and the answers I'm getting are
> > > below.  I don't know if
> > > I'm doing something wrong, or if there is a
> > problem
> > > with the method.
> > > When I do the problem by hand, I get the results
> > > from the book. I've
> > > tried several examples from the book, and none
> of
> > > them agree with what
> > > I'm getting as output from the method.
> > >
> > >
> > > // A short amount of test code
> > > double [][] info =
> {{1,2,-3},{-3,-4,13},{2,1,-5}};
> > > DoubleSquareMatrix dmat = new
> > > DoubleSquareMatrix(info);
> > > System.out.println("dmat="+dmat);
> > > AbstractDoubleSquareMatrix [] res =
> > > dmat.luDecompose(null);
> > > System.out.println("L="+res[0]);
> > > System.out.println("U="+res[1]);
> > >
> > > //resulted in the following output
> > >
> > > dmat=1.0 2.0 -3.0
> > > -3.0 -4.0 13.0
> > > 2.0 1.0 -5.0
> > >
> > > L=1.0 0.0 0.0
> > > -0.6666666666666666 1.0 0.0
> > > -0.3333333333333333 -0.4000000000000001 1.0
> > >
> > > U=-3.0 -4.0 13.0
> > > 0.0 -1.6666666666666665 3.666666666666666
> > > 0.0 0.0 2.8
> > >
> > > Any help is appreciated.
> > > Thanks,
> > > Steve Sinai
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > > Yahoo! Groups Links
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
> > >
>
=== message truncated ===


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Added v0.944.

----- Original Message ----
From: Steve Sinai <ssinai@...>
To: jsci@yahoogroups.com
Sent: Wednesday, 31 May, 2006 12:10:18 AM
Subject: Re: [jsci] luDecompose problem in DoubleSquareMatrix

Hi Mark,

If it isn't too much trouble, adding the textbook
version of LU decomposition to JSci would be helpful.
Part of the reason for my editor is as a teaching aid,
so even if the "classic LU decomposition" (as my book
describes it) isn't used much outside of the
classroom, in this case the classroom version would be
nice.

I guess the phrase "classic LU decomposition" is a
polite way of saying that it's not too useful for
modern, real-world applications.

Thanks,
Steve Sinai

--- Mark Hale <mj_hale@...> wrote:

> Yes, it does.  The LU decomposition is often an
> intermediary step, so the fact that it is permutated
> (scrambled) usually doesn't matter because you
> either unscramble at a later step (e.g. matrix
> inversion), or the final calculation isn't affected
> (so much) by permutations (e.g. matrix determinant).
>  If you are calculating the LU just for its own sake
> (e.g. textbook) then you need an LU method without
> pivoting...which I can provide if desired.
>
> ----- Original Message ----
> From: Steve Sinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Tuesday, 30 May, 2006 10:01:28 AM
> Subject: Re: [jsci] luDecompose problem in
> DoubleSquareMatrix
>
> Hi Mark,
>
> I'm always impressed with your quick responses.
>
> Unfortunately, I'm one of those people who hit the
> math wall at long division, and have no idea what
> matrices have to do with permutations. I see it's
> mentioned later in the linear algebra book I'm
> slogging through, and I'll eventually get to it. No
> need for you to try explaining it to me.  It would
> be
> futile.
>
> Anyway, I've never actually used an LU decomposition
> for anything.  I'm just trying to write this weird
> text-based math editor (like MATLAB or MAPLE) using
> JSci, and I'm reading through a linear algebra book
> and testing out the editor using the examples in the
> book.  Does the JSci implementation of LU
> Decomposition reflect how people in the real world
> use
> it? If so, I'll just leave it as is.
>
> Thanks,
> Steve Sinai
>
> --- Mark Hale <mj_hale@...> wrote:
>
> > Hi,
> >
> >  luDecompose performs the LU decomposition up to a
> > permutation given by the pivot array.  That is,
> the
> > original matrix M is permutated to a matrix M' and
> > then the LU decomposition is performed on M'.  M'
> is
> > chosen to improve the numerical stability. You can
> > use the pivot array to transform back the output
> but
> > ofcourse it won't be lower triangular any more...
> >
> >  If you disable the pivot code in the source, you
> > should get what you want (but more prone to
> > numerical instabilities).
> >
> >  Regards,
> >
> >  Mark
> >
> > ----- Original Message ----
> > From: ssinai <ssinai@...>
> > To: jsci@yahoogroups.com
> > Sent: Monday, 29 May, 2006 10:10:05 PM
> > Subject: [jsci] luDecompose problem in
> > DoubleSquareMatrix
> >
> > Hi,
> >
> > I'm trying to use the luDecompose method of
> > DoubleSquareMatrix, and am
> > not getting the correct answer to my test
> problems.
> > I'm hardly a
> > master of LU Decomposition, and am using some
> > example problems from a
> > linear algebra text.  Anyway, the initial matrix
> > from the text is
> >  1  2 -3
> > -3 -4 13
> >  2  1 -5
> > and the answer the text gives is
> > L= 1    0  0
> >   -3    1  0
> >    2 -3/2  1
> >
> > U= 1 2 -3
> >    0 2  4
> >    0 0  7
> >
> > Some test code and the answers I'm getting are
> > below.  I don't know if
> > I'm doing something wrong, or if there is a
> problem
> > with the method.
> > When I do the problem by hand, I get the results
> > from the book. I've
> > tried several examples from the book, and none of
> > them agree with what
> > I'm getting as output from the method.
> >
> >
> > // A short amount of test code
> > double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
> > DoubleSquareMatrix dmat = new
> > DoubleSquareMatrix(info);
> > System.out.println("dmat="+dmat);
> > AbstractDoubleSquareMatrix [] res =
> > dmat.luDecompose(null);
> > System.out.println("L="+res[0]);
> > System.out.println("U="+res[1]);
> >
> > //resulted in the following output
> >
> > dmat=1.0 2.0 -3.0
> > -3.0 -4.0 13.0
> > 2.0 1.0 -5.0
> >
> > L=1.0 0.0 0.0
> > -0.6666666666666666 1.0 0.0
> > -0.3333333333333333 -0.4000000000000001 1.0
> >
> > U=-3.0 -4.0 13.0
> > 0.0 -1.6666666666666665 3.666666666666666
> > 0.0 0.0 2.8
> >
> > Any help is appreciated.
> > Thanks,
> > Steve Sinai
> >
> >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam
> protection around
> http://mail.yahoo.com
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com




Yahoo! Groups Links

Hi Mark,

If it isn't too much trouble, adding the textbook
version of LU decomposition to JSci would be helpful.
Part of the reason for my editor is as a teaching aid,
so even if the "classic LU decomposition" (as my book
describes it) isn't used much outside of the
classroom, in this case the classroom version would be
nice.

I guess the phrase "classic LU decomposition" is a
polite way of saying that it's not too useful for
modern, real-world applications.

Thanks,
Steve Sinai

--- Mark Hale <mj_hale@...> wrote:

> Yes, it does.  The LU decomposition is often an
> intermediary step, so the fact that it is permutated
> (scrambled) usually doesn't matter because you
> either unscramble at a later step (e.g. matrix
> inversion), or the final calculation isn't affected
> (so much) by permutations (e.g. matrix determinant).
>  If you are calculating the LU just for its own sake
> (e.g. textbook) then you need an LU method without
> pivoting...which I can provide if desired.
>
> ----- Original Message ----
> From: Steve Sinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Tuesday, 30 May, 2006 10:01:28 AM
> Subject: Re: [jsci] luDecompose problem in
> DoubleSquareMatrix
>
> Hi Mark,
>
> I'm always impressed with your quick responses.
>
> Unfortunately, I'm one of those people who hit the
> math wall at long division, and have no idea what
> matrices have to do with permutations. I see it's
> mentioned later in the linear algebra book I'm
> slogging through, and I'll eventually get to it. No
> need for you to try explaining it to me.  It would
> be
> futile.
>
> Anyway, I've never actually used an LU decomposition
> for anything.  I'm just trying to write this weird
> text-based math editor (like MATLAB or MAPLE) using
> JSci, and I'm reading through a linear algebra book
> and testing out the editor using the examples in the
> book.  Does the JSci implementation of LU
> Decomposition reflect how people in the real world
> use
> it? If so, I'll just leave it as is.
>
> Thanks,
> Steve Sinai
>
> --- Mark Hale <mj_hale@...> wrote:
>
> > Hi,
> >
> >  luDecompose performs the LU decomposition up to a
> > permutation given by the pivot array.  That is,
> the
> > original matrix M is permutated to a matrix M' and
> > then the LU decomposition is performed on M'.  M'
> is
> > chosen to improve the numerical stability. You can
> > use the pivot array to transform back the output
> but
> > ofcourse it won't be lower triangular any more...
> >
> >  If you disable the pivot code in the source, you
> > should get what you want (but more prone to
> > numerical instabilities).
> >
> >  Regards,
> >
> >  Mark
> >
> > ----- Original Message ----
> > From: ssinai <ssinai@...>
> > To: jsci@yahoogroups.com
> > Sent: Monday, 29 May, 2006 10:10:05 PM
> > Subject: [jsci] luDecompose problem in
> > DoubleSquareMatrix
> >
> > Hi,
> >
> > I'm trying to use the luDecompose method of
> > DoubleSquareMatrix, and am
> > not getting the correct answer to my test
> problems.
> > I'm hardly a
> > master of LU Decomposition, and am using some
> > example problems from a
> > linear algebra text.  Anyway, the initial matrix
> > from the text is
> >  1  2 -3
> > -3 -4 13
> >  2  1 -5
> > and the answer the text gives is
> > L= 1    0  0
> >   -3    1  0
> >    2 -3/2  1
> >
> > U= 1 2 -3
> >    0 2  4
> >    0 0  7
> >
> > Some test code and the answers I'm getting are
> > below.  I don't know if
> > I'm doing something wrong, or if there is a
> problem
> > with the method.
> > When I do the problem by hand, I get the results
> > from the book. I've
> > tried several examples from the book, and none of
> > them agree with what
> > I'm getting as output from the method.
> >
> >
> > // A short amount of test code
> > double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
> > DoubleSquareMatrix dmat = new
> > DoubleSquareMatrix(info);
> > System.out.println("dmat="+dmat);
> > AbstractDoubleSquareMatrix [] res =
> > dmat.luDecompose(null);
> > System.out.println("L="+res[0]);
> > System.out.println("U="+res[1]);
> >
> > //resulted in the following output
> >
> > dmat=1.0 2.0 -3.0
> > -3.0 -4.0 13.0
> > 2.0 1.0 -5.0
> >
> > L=1.0 0.0 0.0
> > -0.6666666666666666 1.0 0.0
> > -0.3333333333333333 -0.4000000000000001 1.0
> >
> > U=-3.0 -4.0 13.0
> > 0.0 -1.6666666666666665 3.666666666666666
> > 0.0 0.0 2.8
> >
> > Any help is appreciated.
> > Thanks,
> > Steve Sinai
> >
> >
> >
> >
> >
> >
> >
> > Yahoo! Groups Links
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
> >
>
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam?  Yahoo! Mail has the best spam
> protection around
> http://mail.yahoo.com
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Yes, it does.  The LU decomposition is often an intermediary step, so the fact
that it is permutated (scrambled) usually doesn't matter because you either
unscramble at a later step (e.g. matrix inversion), or the final calculation
isn't affected (so much) by permutations (e.g. matrix determinant).  If you are
calculating the LU just for its own sake (e.g. textbook) then you need an LU
method without pivoting...which I can provide if desired.

----- Original Message ----
From: Steve Sinai <ssinai@...>
To: jsci@yahoogroups.com
Sent: Tuesday, 30 May, 2006 10:01:28 AM
Subject: Re: [jsci] luDecompose problem in DoubleSquareMatrix

Hi Mark,

I'm always impressed with your quick responses.

Unfortunately, I'm one of those people who hit the
math wall at long division, and have no idea what
matrices have to do with permutations. I see it's
mentioned later in the linear algebra book I'm
slogging through, and I'll eventually get to it. No
need for you to try explaining it to me.  It would be
futile.

Anyway, I've never actually used an LU decomposition
for anything.  I'm just trying to write this weird
text-based math editor (like MATLAB or MAPLE) using
JSci, and I'm reading through a linear algebra book
and testing out the editor using the examples in the
book.  Does the JSci implementation of LU
Decomposition reflect how people in the real world use
it? If so, I'll just leave it as is.

Thanks,
Steve Sinai

--- Mark Hale <mj_hale@...> wrote:

> Hi,
>
>  luDecompose performs the LU decomposition up to a
> permutation given by the pivot array.  That is, the
> original matrix M is permutated to a matrix M' and
> then the LU decomposition is performed on M'.  M' is
> chosen to improve the numerical stability. You can
> use the pivot array to transform back the output but
> ofcourse it won't be lower triangular any more...
>
>  If you disable the pivot code in the source, you
> should get what you want (but more prone to
> numerical instabilities).
>
>  Regards,
>
>  Mark
>
> ----- Original Message ----
> From: ssinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Monday, 29 May, 2006 10:10:05 PM
> Subject: [jsci] luDecompose problem in
> DoubleSquareMatrix
>
> Hi,
>
> I'm trying to use the luDecompose method of
> DoubleSquareMatrix, and am
> not getting the correct answer to my test problems.
> I'm hardly a
> master of LU Decomposition, and am using some
> example problems from a
> linear algebra text.  Anyway, the initial matrix
> from the text is
>  1  2 -3
> -3 -4 13
>  2  1 -5
> and the answer the text gives is
> L= 1    0  0
>   -3    1  0
>    2 -3/2  1
>
> U= 1 2 -3
>    0 2  4
>    0 0  7
>
> Some test code and the answers I'm getting are
> below.  I don't know if
> I'm doing something wrong, or if there is a problem
> with the method.
> When I do the problem by hand, I get the results
> from the book. I've
> tried several examples from the book, and none of
> them agree with what
> I'm getting as output from the method.
>
>
> // A short amount of test code
> double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
> DoubleSquareMatrix dmat = new
> DoubleSquareMatrix(info);
> System.out.println("dmat="+dmat);
> AbstractDoubleSquareMatrix [] res =
> dmat.luDecompose(null);
> System.out.println("L="+res[0]);
> System.out.println("U="+res[1]);
>
> //resulted in the following output
>
> dmat=1.0 2.0 -3.0
> -3.0 -4.0 13.0
> 2.0 1.0 -5.0
>
> L=1.0 0.0 0.0
> -0.6666666666666666 1.0 0.0
> -0.3333333333333333 -0.4000000000000001 1.0
>
> U=-3.0 -4.0 13.0
> 0.0 -1.6666666666666665 3.666666666666666
> 0.0 0.0 2.8
>
> Any help is appreciated.
> Thanks,
> Steve Sinai
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com




Yahoo! Groups Links

Hi Mark,

I'm always impressed with your quick responses.

Unfortunately, I'm one of those people who hit the
math wall at long division, and have no idea what
matrices have to do with permutations. I see it's
mentioned later in the linear algebra book I'm
slogging through, and I'll eventually get to it. No
need for you to try explaining it to me.  It would be
futile.

Anyway, I've never actually used an LU decomposition
for anything.  I'm just trying to write this weird
text-based math editor (like MATLAB or MAPLE) using
JSci, and I'm reading through a linear algebra book
and testing out the editor using the examples in the
book.  Does the JSci implementation of LU
Decomposition reflect how people in the real world use
it? If so, I'll just leave it as is.

Thanks,
Steve Sinai

--- Mark Hale <mj_hale@...> wrote:

> Hi,
>
>  luDecompose performs the LU decomposition up to a
> permutation given by the pivot array.  That is, the
> original matrix M is permutated to a matrix M' and
> then the LU decomposition is performed on M'.  M' is
> chosen to improve the numerical stability. You can
> use the pivot array to transform back the output but
> ofcourse it won't be lower triangular any more...
>
>  If you disable the pivot code in the source, you
> should get what you want (but more prone to
> numerical instabilities).
>
>  Regards,
>
>  Mark
>
> ----- Original Message ----
> From: ssinai <ssinai@...>
> To: jsci@yahoogroups.com
> Sent: Monday, 29 May, 2006 10:10:05 PM
> Subject: [jsci] luDecompose problem in
> DoubleSquareMatrix
>
> Hi,
>
> I'm trying to use the luDecompose method of
> DoubleSquareMatrix, and am
> not getting the correct answer to my test problems.
> I'm hardly a
> master of LU Decomposition, and am using some
> example problems from a
> linear algebra text.  Anyway, the initial matrix
> from the text is
>  1  2 -3
> -3 -4 13
>  2  1 -5
> and the answer the text gives is
> L= 1    0  0
>   -3    1  0
>    2 -3/2  1
>
> U= 1 2 -3
>    0 2  4
>    0 0  7
>
> Some test code and the answers I'm getting are
> below.  I don't know if
> I'm doing something wrong, or if there is a problem
> with the method.
> When I do the problem by hand, I get the results
> from the book. I've
> tried several examples from the book, and none of
> them agree with what
> I'm getting as output from the method.
>
>
> // A short amount of test code
> double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
> DoubleSquareMatrix dmat = new
> DoubleSquareMatrix(info);
> System.out.println("dmat="+dmat);
> AbstractDoubleSquareMatrix [] res =
> dmat.luDecompose(null);
> System.out.println("L="+res[0]);
> System.out.println("U="+res[1]);
>
> //resulted in the following output
>
> dmat=1.0 2.0 -3.0
> -3.0 -4.0 13.0
> 2.0 1.0 -5.0
>
> L=1.0 0.0 0.0
> -0.6666666666666666 1.0 0.0
> -0.3333333333333333 -0.4000000000000001 1.0
>
> U=-3.0 -4.0 13.0
> 0.0 -1.6666666666666665 3.666666666666666
> 0.0 0.0 2.8
>
> Any help is appreciated.
> Thanks,
> Steve Sinai
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>
>
>
>
>
>
>
>
>


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around
http://mail.yahoo.com

Hi,

  luDecompose performs the LU decomposition up to a permutation given by the
pivot array.  That is, the original matrix M is permutated to a matrix M' and
then the LU decomposition is performed on M'.  M' is chosen to improve the
numerical stability. You can use the pivot array to transform back the output
but ofcourse it won't be lower triangular any more...

  If you disable the pivot code in the source, you should get what you want (but
more prone to numerical instabilities).

  Regards,

  Mark

----- Original Message ----
From: ssinai <ssinai@...>
To: jsci@yahoogroups.com
Sent: Monday, 29 May, 2006 10:10:05 PM
Subject: [jsci] luDecompose problem in DoubleSquareMatrix

Hi,

I'm trying to use the luDecompose method of DoubleSquareMatrix, and am
not getting the correct answer to my test problems.  I'm hardly a
master of LU Decomposition, and am using some example problems from a
linear algebra text.  Anyway, the initial matrix from the text is
  1  2 -3
-3 -4 13
  2  1 -5
and the answer the text gives is
L= 1    0  0
   -3    1  0
    2 -3/2  1

U= 1 2 -3
    0 2  4
    0 0  7

Some test code and the answers I'm getting are below.  I don't know if
I'm doing something wrong, or if there is a problem with the method.
When I do the problem by hand, I get the results from the book. I've
tried several examples from the book, and none of them agree with what
I'm getting as output from the method.


// A short amount of test code
double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
DoubleSquareMatrix dmat = new DoubleSquareMatrix(info);
System.out.println("dmat="+dmat);
AbstractDoubleSquareMatrix [] res = dmat.luDecompose(null);
System.out.println("L="+res[0]);
System.out.println("U="+res[1]);

//resulted in the following output

dmat=1.0 2.0 -3.0
-3.0 -4.0 13.0
2.0 1.0 -5.0

L=1.0 0.0 0.0
-0.6666666666666666 1.0 0.0
-0.3333333333333333 -0.4000000000000001 1.0

U=-3.0 -4.0 13.0
0.0 -1.6666666666666665 3.666666666666666
0.0 0.0 2.8

Any help is appreciated.
Thanks,
Steve Sinai







Yahoo! Groups Links

Hi,

I'm trying to use the luDecompose method of DoubleSquareMatrix, and am
not getting the correct answer to my test problems.  I'm hardly a
master of LU Decomposition, and am using some example problems from a
linear algebra text.  Anyway, the initial matrix from the text is
  1  2 -3
-3 -4 13
  2  1 -5
and the answer the text gives is
L= 1    0  0
   -3    1  0
    2 -3/2  1

U= 1 2 -3
    0 2  4
    0 0  7

Some test code and the answers I'm getting are below.  I don't know if
I'm doing something wrong, or if there is a problem with the method.
When I do the problem by hand, I get the results from the book. I've
tried several examples from the book, and none of them agree with what
I'm getting as output from the method.


// A short amount of test code
double [][] info = {{1,2,-3},{-3,-4,13},{2,1,-5}};
DoubleSquareMatrix dmat = new DoubleSquareMatrix(info);
System.out.println("dmat="+dmat);
AbstractDoubleSquareMatrix [] res = dmat.luDecompose(null);
System.out.println("L="+res[0]);
System.out.println("U="+res[1]);

//resulted in the following output

dmat=1.0 2.0 -3.0
-3.0 -4.0 13.0
2.0 1.0 -5.0

L=1.0 0.0 0.0
-0.6666666666666666 1.0 0.0
-0.3333333333333333 -0.4000000000000001 1.0

U=-3.0 -4.0 13.0
0.0 -1.6666666666666665 3.666666666666666
0.0 0.0 2.8

Any help is appreciated.
Thanks,
Steve Sinai

Ahhh, I tested EngineerMath.resample(), not
signal.resample().

Are you doing out=signal.resample().evaluate()?
Because that would involve Cascades and that seems to
apply a low pass filter to the data, which might
explain the difference in your graphs.

--- simon_982 <no_reply@yahoogroups.com> wrote:

> --- In jsci@yahoogroups.com, Mark Hale <mj_hale@...>
> wrote:
> >
> > I've added a test case for resample() to the cvs.
> It
> > looks like it works ok.
>
> Sorry I didn't understand where u added the test
> case to...
> Anyway if u look at the images i posted u can see
> that the two signal
> are quite different.
> The only thing i did was to use the
> signal.resample(N) method where
> signal was the signal to be resampled and N was the
> number of samples
> i wanted to obtain.
> I don't think I did it wrong...
>
> Always thanks a lot for your support Mark!
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>
>




___________________________________________________________
To help you stay safe and secure online, we've developed the all new Yahoo!
Security Centre. http://uk.security.yahoo.com

My thinking would be something like this: (psuedocode)
signal(N)
signalPadded(M) =  pad(signal)
output(M-F+1) = signalPadded(M).convolve(filter(F))
compare(signal, output)

so if you pick the length M of the padded signal and
the length F of the filter to be such that M-F=N-1,
then the output should be of the same length as the
original signal. If M-F > N-1 then I would assume you
would just get a padded output signal, which you could
then trim back to N.


--- simon_982 <no_reply@yahoogroups.com> wrote:

> --- In jsci@yahoogroups.com, Mark Hale <mj_hale@...>
> wrote:
> >
> > If you have a signal of 2N and filter of N, you
> should
> > get an output of N+1.
>
> And that is the correct size of the result? What if
> i want to compare
> it with the original signal? Should I resample it to
> 2N samples? Or
> that is not correct?
>
>
> > Is there any known standard/example signal that
> when
> > convoluted with some std/exmpl filter gives a std
> > result that I could use to build a test case?
> > (e.g. for Fourier transforms, the FT of a guassian
> is
> > a guassian). I am not too familar with some of
> this
> > stuff.
>
> Sorry I don't know this either... :(
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>
>




___________________________________________________________
To help you stay safe and secure online, we've developed the all new Yahoo!
Security Centre. http://uk.security.yahoo.com

--- In jsci@yahoogroups.com, Mark Hale <mj_hale@...> wrote:
>
> I've added a test case for resample() to the cvs. It
> looks like it works ok.

Sorry I didn't understand where u added the test case to...
Anyway if u look at the images i posted u can see that the two signal
are quite different.
The only thing i did was to use the signal.resample(N) method where
signal was the signal to be resampled and N was the number of samples
i wanted to obtain.
I don't think I did it wrong...

Always thanks a lot for your support Mark!

--- In jsci@yahoogroups.com, Mark Hale <mj_hale@...> wrote:
>
> If you have a signal of 2N and filter of N, you should
> get an output of N+1.

And that is the correct size of the result? What if i want to compare
it with the original signal? Should I resample it to 2N samples? Or
that is not correct?


> Is there any known standard/example signal that when
> convoluted with some std/exmpl filter gives a std
> result that I could use to build a test case?
> (e.g. for Fourier transforms, the FT of a guassian is
> a guassian). I am not too familar with some of this
> stuff.

Sorry I don't know this either... :(

I've added a test case for resample() to the cvs. It
looks like it works ok.

--- simon_982 <no_reply@yahoogroups.com> wrote:

> In addition to what i just wrote, why does the
> resample method return
> me a distorted and noisy signal?!?
> Is there anything wrong with the java code or am i
> misunderstanding
> anything?
>
> Look here the original signal:
> http://digilander.libero.it/friendsita/Original.gif
>
> And the resampled one:
> http://digilander.libero.it/friendsita/Resampled.gif
>
> Could seem a small problem but gives me real
> troubles with wavelet
> transforms.
>
> Bye
> Simon
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>




___________________________________________________________
To help you stay safe and secure online, we've developed the all new Yahoo!
Security Centre. http://uk.security.yahoo.com

In addition to what i just wrote, why does the resample method return
me a distorted and noisy signal?!?
Is there anything wrong with the java code or am i misunderstanding
anything?

Look here the original signal:
http://digilander.libero.it/friendsita/Original.gif

And the resampled one:
http://digilander.libero.it/friendsita/Resampled.gif

Could seem a small problem but gives me real troubles with wavelet
transforms.

Bye
Simon

If you have a signal of 2N and filter of N, you should
get an output of N+1.

Is there any known standard/example signal that when
convoluted with some std/exmpl filter gives a std
result that I could use to build a test case?
(e.g. for Fourier transforms, the FT of a guassian is
a guassian). I am not too familar with some of this
stuff.

--- simon_982 <no_reply@yahoogroups.com> wrote:

> --- In jsci@yahoogroups.com, Mark Hale <mj_hale@...>
> wrote:
> >
> > It has to, otherwise the array index would go out
> of
> > bounds:
> >
> > y(k) = \sum_{l=0}^m f(l) x(k-l)
> >
> > In the discrete case, signals only extend over a
> > finite range. What you need to do is pad the
> original
> > signal, either with zeros (linear convolution) or
> > extend it periodically (cyclic convolution).
>
> I've unsterstood the reason why it would go out of
> bound.
> Anyway I extended the original signal simmetrically
> (to avoid
> discontinuity problems) with a total of 2N samples
> (with N the
> original number of samples) and applied again the
> same filters as
> before.
> Every time i obtain a shorter signal as i use a
> larger filter.
> So the problem didn't solve...
>
> Anyway, if the signal is shorter in its number of
> samples, does this
> mean that the result of the convolution is
> truncated? Or is it right?
> Could i use linear interpolation to obtain the
> number of samples i
> need without changing the nature of the convolution
> result?
>
> I hope I expressed my doubts clearly...
>
>
> > I notice the actual Java code does
> > y(k) = \sum_{l=0}^m f(l) x(k+l)
> > not sure whether the k+l is a bug or choice of
> > convention...
>
> Could anyone check this issue please?
>
> Thanks a lot, your help is great.
> Simon
>
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>
>




___________________________________________________________
Win a BlackBerry device from O2 with Yahoo!. Enter now.
http://www.yahoo.co.uk/blackberry

--- In jsci@yahoogroups.com, Mark Hale <mj_hale@...> wrote:
>
> It has to, otherwise the array index would go out of
> bounds:
>
> y(k) = \sum_{l=0}^m f(l) x(k-l)
>
> In the discrete case, signals only extend over a
> finite range. What you need to do is pad the original
> signal, either with zeros (linear convolution) or
> extend it periodically (cyclic convolution).

I've unsterstood the reason why it would go out of bound.
Anyway I extended the original signal simmetrically (to avoid
discontinuity problems) with a total of 2N samples (with N the
original number of samples) and applied again the same filters as
before.
Every time i obtain a shorter signal as i use a larger filter.
So the problem didn't solve...

Anyway, if the signal is shorter in its number of samples, does this
mean that the result of the convolution is truncated? Or is it right?
Could i use linear interpolation to obtain the number of samples i
need without changing the nature of the convolution result?

I hope I expressed my doubts clearly...


> I notice the actual Java code does
> y(k) = \sum_{l=0}^m f(l) x(k+l)
> not sure whether the k+l is a bug or choice of
> convention...

Could anyone check this issue please?

Thanks a lot, your help is great.
Simon

It has to, otherwise the array index would go out of
bounds:

y(k) = \sum_{l=0}^m f(l) x(k-l)

In the discrete case, signals only extend over a
finite range. What you need to do is pad the original
signal, either with zeros (linear convolution) or
extend it periodically (cyclic convolution).

I notice the actual Java code does
y(k) = \sum_{l=0}^m f(l) x(k+l)
not sure whether the k+l is a bug or choice of
convention...


--- simon_982 <no_reply@yahoogroups.com> wrote:

> I was wondering why the method
> ____
>  Signal filter(double[] f)
>           Apply the given array as a convolution
> Filter and return a
> new Signal.
> ____
>
> does return a shorter (in number of samples) signal
> according to the
> size of the applied filter vector.
> I need to have the resulting signal always at the
> same size
> independently of the filter size.
>
> Is this possible?
> Simon
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>
>






___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail
http://uk.messenger.yahoo.com

I was wondering why the method
____
  Signal filter(double[] f)
           Apply the given array as a convolution Filter and return a
new Signal.
____

does return a shorter (in number of samples) signal according to the
size of the applied filter vector.
I need to have the resulting signal always at the same size
independently of the filter size.

Is this possible?
Simon

Unfortunately, I know even less about wavelets :/ I
believe the DougJamesDau2 example is similar to what
you want to do. But use the
JSci.maths.wavelets.splines.QuadraticSpline class
instead of Daubechies2. There maybe boundary padding
issues for high levels...


--- simon_982 <no_reply@yahoogroups.com> wrote:

> Ok, now that i solved some silly problem i really
> need your help.
> I've got a discrete signal stored in a 256 elements
> array.
> What i need is to compute its dyadic wavelet, the
> quadratic spline
> of compact support defined below, at four different
> resolution
> levels for 3<=j<=6.
>
> The quadratic spline of compact support that i need
> is the one
> mentioned at page 4 of this pdf (
>
http://www.cmap.polytechnique.fr/~mallat/papiers/MallatEdgeCharact92.
> pdf ) and is further defined at page 19 with its
> filters H, G, K, L.
>
> Now, i know really few things about wavelets, i just
> need to compute
> this one. Can i use this JSci package to do this?
> Could anyone
> please gimme a detailed description of what I should
> do?
>
> Or maybe could i simply use the algorithm described
> at page 20?
>
> Please, any help would be much appreciated.
> Thanks
> Simon
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>






___________________________________________________________
Yahoo! Messenger - NEW crystal clear PC to PC calling worldwide with voicemail
http://uk.messenger.yahoo.com

Ok, now that i solved some silly problem i really need your help.
I've got a discrete signal stored in a 256 elements array.
What i need is to compute its dyadic wavelet, the quadratic spline
of compact support defined below, at four different resolution
levels for 3<=j<=6.

The quadratic spline of compact support that i need is the one
mentioned at page 4 of this pdf (
http://www.cmap.polytechnique.fr/~mallat/papiers/MallatEdgeCharact92.
pdf ) and is further defined at page 19 with its filters H, G, K, L.

Now, i know really few things about wavelets, i just need to compute
this one. Can i use this JSci package to do this? Could anyone
please gimme a detailed description of what I should do?

Or maybe could i simply use the algorithm described at page 20?

Please, any help would be much appreciated.
Thanks
Simon

--- In jsci@yahoogroups.com, Mark Hale <mj_hale@...> wrote:
>
> Try
> import JSci.maths.*;

Thanks a lot, that worked.
I can't believe it was only that, it was drivin me crazy...

Try
import JSci.maths.*;


--- simon_982 <no_reply@yahoogroups.com> wrote:

> Hi everyone, i'm having hard times trying to get the
> JSci package to
> work.
> I've literally followed the instructions contained
> in the index.html
> file:
> 1- unzipped the JSci.zip file to
> G:\Programmi\Java\JSci
> 2- since i'm on WinXP, opened Control Panel, System,
> Advanced,
> Enviroment variables, added the classpath
> G:\Programmi\Java\JSci\classes\jsci-core.jar to the
> current
> classpath variable belonging to my userid.
> 3- from msdos prompt tried to compile a generic
> class containing
> import JSci.maths;
>
>
> I always get the following error:
>
> class.java:8: cannot resolve symbol
> symbol  : class maths
> location: package JSci
> import JSci.maths;
>             ^
> 1 error
>
> The same thing happens also for the other package i
> need, which is
> the wavelet one, even if i've set the right
> classpath for it too.
>
> Please help me cause i really need to use this
> package for my thesys
> and this is the first time i'm installing a java
> package.
> Thanks in advance
> Simon
>
>
>
>
>
>
> Yahoo! Groups Links
>
>
>     jsci-unsubscribe@yahoogroups.com
>
>
>
>
>
>




___________________________________________________________
To help you stay safe and secure online, we've developed the all new Yahoo!
Security Centre. http://uk.security.yahoo.com

Hi everyone, i'm having hard times trying to get the JSci package to
work.
I've literally followed the instructions contained in the index.html
file:
1- unzipped the JSci.zip file to G:\Programmi\Java\JSci
2- since i'm on WinXP, opened Control Panel, System, Advanced,
Enviroment variables, added the classpath
G:\Programmi\Java\JSci\classes\jsci-core.jar to the current
classpath variable belonging to my userid.
3- from msdos prompt tried to compile a generic class containing
import JSci.maths;


I always get the following error:

class.java:8: cannot resolve symbol
symbol  : class maths
location: package JSci
import JSci.maths;
             ^
1 error

The same thing happens also for the other package i need, which is
the wavelet one, even if i've set the right classpath for it too.

Please help me cause i really need to use this package for my thesys
and this is the first time i'm installing a java package.
Thanks in advance
Simon

Hi Mark,
>  I've committed to the CVS an updated MathMLParser
>  class for you. I have added static translateXXX(Node
>  node) methods. You can pass them the MathMLDocument
>  node to get all the representations.
many thanks for the really fast help.

Regards
      Torsten



___________________________________________________________
Telefonate ohne weitere Kosten vom PC zum PC: http://messenger.yahoo.de