Welcome to the Netlib Discussion Forum. This forum is divided into topics, each of which may contain subtopics. Browse through the topics by clicking on the "Topics" link in the left frame to read and/or post discussions. If you are a new user, you may want to read the Frequently Asked Questions by clicking on the "FAQ" link in the left frame. Also, notice that you can format your messages easily; the "Formatting" link on the left gives the details. Enjoy!!!

HI,

I want to solve a set of nonlinear equations, g1(x1,x2,...,xn),...., gn(x1,x2,...,xn), and n can be very large (I am trying 70 for my current project), and I find that under some cases, there are multiple roots for the above set of nonlinear equations, (this leads to the "jump phenomea", and can you tell me what program and what software can get MULTIPLE roots of the above equation instead of just one root.

Meanwhile, I am using the hybrj.f to implement my computation.

thanks

I wonder if anybody knows about industrial
applications in commercial FE software
of preconditioned iterative
methods based on Schur Complement for
independent set orderings.

Hi,

I want to get a 3D contour(line & fill) algorithm for a single triangular & quadratic mesh to visualize computed stress result on each vertex of that element. I found some Fortran sources here, but these seemed so complicated to me and these are only for 2D. I prefer C-language style code, but Fortran is also OK. Any comments are to be wellcomed from anyone knows about this algorithm.

thanks

Hi
I want to get a program (software) to minimize
HIGHLY NONLIEAR MULTIVARIATE objective function
with linear constraints.
I will appreciate if anybody helps me regarding
this.

Thanking you

(Alak K. Datta)

Hi
I am looking for a FORTRAN routine for solving the equation which looks something like this:

g11 = f(x,theta,phi); g12 = f1(x,theta,phi) ....
where g11... are the elements of a 3 x 3 matrix.

x is complex.

f = Double integral with respect to "theta" and "phi" with appropriate limits.

The aim is to determine x to satisfy |G| = 0

G = DET[g11,g12,g13; g21,g22,g23;g31,g32,g33] == 0

I would be thankful if anybody could help me regarding this

Thanking you in anticipation

V. Munikoti

i need the complexity for computing the inverse of
a real symmetric matrix using a single processor

please e-mail me if you have an answer

I would like to find out an algorithm that fits the ODE system to an experimental data...

Hi all

I am currently trying to perform SVD of a complex
general 3x3 matrix using the fortran subroutine
CGESVD.F. Comments within the subroutine state
that on output VT is in fact V**H, rather than
just the matrix V. Has anyone used this code
before and have an idea as to what H is!? I need
to recover the matrix V.

Thanks in advance ....

F77 Compatible BLAS vs F90 builtin linear algebra

Do BLAS routines exist for F90? If not, how do the built in F90 linear algebra routines (eg matmul) compare with calling the F77 BLAS routines.

Thanks

I have BLAS routines written in a subset of F90 at
my web site (http://www.ozemail.com.au/~milleraj).
They make some use of F90 array syntax.

I would be interested particularly in comments on
routine DNRM2 in level-1 BLAS.

Until recently, my version was based upon Hammarling's code, but I have reverted back to
Lawson's original after finding it much faster
for all 4 compilers on which I tested it.

Hi,
My research involves fitting a circle or ellipse in 3D space by means of least squares method on measured datas in the form ( x, y ,z). Does anybody know where I can get the algorithms or source code regarding this. Thanks.

Hi,

I am looking for some information on
Boeing Harwell Sparse matrix Format. I need
to understand what exactly is this format.
I need to represent a graph/grid with verticies
and edges in a BH format.

Any assisntance or references to such information
is greatly appreciated.

Thanks,
priti

Hi, does anyone know the algorithm used by
passport readers?

Thanks!
bill_allen9@hotmail.com

test {attach}

I am a newbie at compiling source codes, etc.

I've download the f2c executable for MS-DOS. After converting I get a file that requires 'libf2c.a'. Where (specifically do I get this file? I've looked all over netlib and is simply overwhelmed. How do I compile into Microsoft C++?

Your help is greatly appreciated.

Hi, All.
I'm currently looking for some mailing list about
phisics. If you know one, please tell me.

Ilia Chipitsine <ilia@cgu.chel.su>

Hi!

I wonder if the documentation to NAPACK
is available anywhere.
A would appreciate type descriptions of
the parameters to the routines of NAPACK,
here on NETLIB.

Thank's

Peter Olivius <petero@matfys.lth.se>

I'd like to configure multiple heterogenous machines (intel, sun ...).One of them will be consedered as the master,the others as slaves. These machines will operate together as a single one to support sequential an parallel programs.
How can I do this ? and what tools do I need ?

Thank's for the help

Error in VARPRO.
VARPRO is very old Fortran 66 code for separable
non-linear least-squares fitting. One important
application is in fitting sums of exponentials.
The time constants/ decay rates are the non-linear
parameters; the amplitudes are the linear ones.

In routine VPDPA there is a variable called SAVE.
As its name suggests, it needs to be saved.
I suspect that most Fortran compilers will cough
and splutter over the statement:
SAVE SAVE
so I suggest that SAVE is renamed as say SAVED,
then you can insert the statement:
SAVE SAVED.

I have a version of VARPRO which is compatible
with Lahey's free ELF90 compiler. My e-mail
address:
Alan.Miller @ vic.cmis.csiro.au

Hi All,

Does anyone have a c version of LSODA routine? I had some difficulties in converting original fortran program to c using f2c. Thank you very much for your help

Hello,

A class of mine requires me to have access to BLAS functions. I have a Windows 98 machine with Visual C++ 6.0.
Where can I get the libraries to access these functions?
How do I install them once I get them?
Fairly detailed directions would be appreciated as I'm new to all of this.

Thank you,
Dennis Lambert

Inverse of a sparse matrix

Does anybody know of any package routines to invert a matrix in a sparse - compressed row - format? are there any in netlib directories?
Thanks.

Does anybody know how to solve a singular equations, Ax=b, where A is a singular spare matrix?

Thanks alot.

Dennis

Levenberg-Maquardt:

What is the LAPACK correspondant of the ZXSSQ function of the ISML libraries?

Several Fortran programs I used to fit NMR relaxation data employed the ZXSSQ routine of the IMSL routine (Vax). Is LAPACK the right package to use to solve Levenberg-Maquardt problems?

Please Email your answer. Thanks:-)

Hi All

I am trying to simulate and control a continuous system with a software programmed in Java.
I am looking for a JAVA routine for solving ODEs. Does anybody know where I can get algorithms or source code in Java ?

Please, Email your answer

Thanks.

hi all

i want to know any program which can be used to
get fourier transform (numerically) for data which
are NOT sampled at a constant sampling frequency.

i have data which is sampled at different times and the times are not an integral multiple of the first.

dhiru

I am using the EISPACK routine for the calculation of the eigen vector and eigen value of a complex general matrix. The routine cg.f and it's dependencies files are used for this. On following the instructions I am getting correct eigen values but not the correct eigen vectors. Please guide me on the issue.

With Regards

im trying to get the documentation reader for slatec to work
docinstall says to compile and link slprep and subsid, well i try to do that but get an error msg
seb1 /var/tmp> f77 *.f
slprep.f:
MAIN slprep:
ifdeck:
ifif:
ifsid:
pscat:
sort:
subsid.f:
cvtcat:
find:
lenstr:
upcase:
Linking:
Undefined first referenced
symbol in file
xermsg_ slprep.o
ld: fatal: Symbol referencing errors. No output written to a.out

i checke in slatec/err to see is there is a routine called xermsg, there is not, so any suggestions?

thanks in advance

Hi:

I am trying to solve Ax=b, A is sparse matrix and huge. I want to use preconditioned biconjugate method. But here I just find program written in Fortran. How and where can I get code in C language.

Tahnks a lot

RE: biconjugate method

I am not aware of a C routine for this method; however, a fortran-to-C converter can be downloaded from Netlib at http://www.netlib.org/f2c/.

Hope this helps.

this is my second attempt to compile slatec stuff
two questsions

1. seb1 /tmp/src> ar cr libslatec.a *.o
internal error: error_message(58)

what does this mean and how do i get rid of it

2. when a compiled slatec stuff earlier into a library and tried to use it i got for example
using the factorial function fac(4) giving me
floating exception as the answer, i dont think
the library compiled properly, is there a way
to test this.

Hello,

I'm interested in solving automatic problem.
So, I have to use the "poles placement" method.

For the moment, I didn't see any routine witch
I can solve my problem.

Could you be so kind to indicate me if a general
Fortran or C routine is available somewhere ?

Tanks you in advance,
Didier Touillon

hi,
i am trying to solve laplaces eqn in a cube using SOR. i am using neumann boundary conditions such that on every face of the cube except the bottom face, the normal derivative is zero. at the bottom face, the normal derivative is a combination of two gaussians (i am trying to simulate a pair of active regions on the solar surface) - one positive, and one negative. my problem is that the iterations dont proceed beyond a certain number which increases as i decrease the error requirement. if anyone has done this kind of thing before could you tell me if there is some easier method? i am checking to see if the relative error at every point goes below a particular value. any suggestions/references are welcome, since i am new at this.
sundar (ohmwhorules@mailcity.com)

Hello,

I would like to invert a symmetric hermitian
matrix of the form

R 0 R 1 R 2 R 3
R-1 R 0 R 1 R 2
R-2 R-1 R 0 R 1
R-3 R-2 R-1 R 0

where R-x is the hermetian of Rx.

Any help would be greatly appreciated

Sorry, the matrix I want to invert

R 0 R 1 R 2 R 3
R-1 R 0 R 1 R 2
R-2 R-1 R 0 R 1
R-3 R-2 R-1 R 0

is only hermitian

I am learning about fourier analysis. As a novice, I have been able to create a power spectrum analysis of a time series or audio signal. And I have can recreate the series from the frequency domain.

My question is how can I use the information on non-periodic data and recreate the wavform's character. For example, project one second in to the future.

I noticed that when I recreate the waveform; the recreated waveform becomes periodic when in fact my original waveform isn't.

Can I use the magnitude and phase angle information to create the non-periodic nature of the waveform?

Or Am I limited to using the base frequency harmonics to reconstruct the waveform?

Lastly, I noticed that when I use the fourier transform information to reconstruct the waveform I get the waveform inverted or it appears graphically 180 degrees out of phase whereas if I use the inverse fourier transform function I can reproduce the waveform precisely.

I don't understand the difference between these two functions. Can someone explain why this occurs?

Hello friends:

For some research work, I need a FORTRAN code for
least square approximation of a function of two variables, say, F(x,y) where the domain of (x,y) is an ellipse with semi-axes (a, b). Therefore, what I need is a FORTRAN code that approximates the function F(x,y) in terms of a polynomial in x and y. Please contact me at mrahman10@hotmail.com or via phone at (248)576-8495. Thank you

I am engaged in solving a non-linear integral equation which as far as I know is the most general fom of a Volterra equation of the second
kind

the form being :

f(x) = g(x) + \int K(x,y)*h(y)dy

Can anyone help? Thanks in advance

Buddhapriya

something I should have added in the previous mail itself is

the x and y are vectors

Hi friends!

I download 3 bessel subroutines (dbesj0.f, dbesk0.f and dbesi0.f) and I used them 3 months before successfully.

When I come back now to verify some of results, the dbesj0.f doesn't work. I always have a "x outside (-1,1)" message given by dcsevl function.

Are there some hints to solve this problem?. May I remind that I am repeating some data which worked before.

Thanks!

Hi ,
I am looking for a routine (preferably in f77) to invert a large sparse non-symmetric complex matrix. Please can anybody help ?

Thank You

hi everybody!!

i have a problem where i have to solve the linear
equation given in the subject. the matrix A is of
1200*1200 size.
i want to know whether i can solve the linear
equation using mathematica/ matlab
pl. note that i am a starter in this package

i would 1st ask u whether the code i write in matlab can be converted into an applet of some such thing, so that any person having access to internet can run my program without any problem.
i would appreciate if u would answer this specifically as the computations which would occur are of the order of 1200*1200 size matrix linear equations and involve solving this for some x times(this number though is not very large)
will the user face problems in execution of the
abovesaid code??
this would help me to get really started on the
problem with MATLAB'S help.


also i did go through the commands which are available thru matlab's help and which r pertaining to sparse matix and also to linear equations.
i am curious about the amount of memory used by the software used to solve a large no of simultaneous equations.
i think that the sparse matrix ( 1200 * 1200), with 1200*7 non-zero elements should first be converted into a banded matrix (of 7 bandwidth) and then into a matrix of size 1200*7 with no zero elements.it is now that the linear system should be solved.

the resultant values should now be used to form
another sparse matrix of same features as above and also the steps henceforth followed.
i would like u to tell me if matlab does this in the background or some more efficient method.

I am using f2c (version 19990503) but am having problems concatenating an implicit length string in a subroutine. I wrote a simple test program to show the error. This is the error f2c gives me.

f2c cprob.f
cprob.f:
MAIN:
catit:
Error on line 33 of cprob.f: impossible element in concatenation

The program works without the concatenation line.
Thank you for any help.


c Program to show that a character variable in a subroutine with an indefinite
c length cannot be concatenated.
c Mark Otto, Biological Statistician work: 301-497-5872
c Population and Habitat Assessment Section fax: 301-497-5871
c Office of Migratory Bird Management e-mail: Mark_Otto@FWS.Gov
c U.S. Fish and Wildlife Service
c Department of the Interior
c 11500 American Holly Drive
c Laurel, Maryland 20708-4016

character first*(10),second*(20)
integer lngth
first='Try it. '
second='You''ll like it.'
lngth=14

c Two strings can be cancatenated in the main routine
c Pass the second to a subroutine

write(6,*)'main: ',first(1:10)//second(1:lngth)
call catit(second(1:14))
stop
end

c Give second an implicit length.
subroutine catit(second)
character first*(10),second*(*)
first='Try it. '

c You can print out the inplicit length string but not cancatenate it.
c Test by commenting out the cancatenation step.
write(6,*)'catit: ',second
c write(6,*)'catit: ',first(1:10)//second
stop
end

Please ignore this message

test

Hi all

Does anybody knows where can i find a good rotine to do neural network modeling ? I am particularly interested in routines using backpropagation method in FORTRAN.

Tx all

Error in file SORREVCOM.F which is part of
templates.

On line 131, there is a GOTO 2
Label 2 is INSIDE an IF block on line 211.

I find it incredible that any compiler would not
give a fatal error. The compilers I have tried
all spluttered at it.

It appears that the END IF should be immediately
before label 2.

Hi everybodY!
If any one knows about the websites related to graph theory or ur doing on this subject please mail me.
I am very need of it.
Expecting reply from you
yours,
hari krishna ravva
ravvas@cyberspace.org

templates - Fortran version.
I now have a list of about 25 serious errors.
These fall into 3 categories:
1. Passing double precision arrays to subroutines
which are expecting integer arrays.
2. Jumping into DO loops from outside.
3. Jumping into IF blocks from outside.

Most compilers can handle the 3rd kind of error
without problems, though some refuse giving a
fatal error message. These have been reported
to: templates @ cs.utk.edu
Until these are fixed, anyone should be extremely
careful in using this code.

Alan Miller
http://www.ozemail.com.au/~amiller

I have a very simple (maybe ridiculous) question ,and i hope someone will help. i have downloaded GNu Fortran g77 and i want to know how to access
the 'slatec' subroutine library. i have already
downloaded slatec into a directory c:slatec, so how do i connect it to g77? do i have to put slatec in my autoexec.bat path? do i make other modifications to autoexec.bat? when using g77, do
i write something like "g77 myprogram.f -slatec"
or what? please somebody help me!
i have a compaq presario PC with Windows 98 and DOS.

I want to solve the sparse least squares
problem Ax approx b, where the m-by-n matrix
A has column full rank.

Q1: What are the most commonly used
direct methods?
Q2. Where can reliable Matlab codes
for direct methods be obtained?
Q3. According to the literature, a
symbolic factorization can be done to
find a permutation matrix P such that
the Cholesky factor of the symm pos def
matrix B = (AP)'*(AP) is as sparse as
possible. Is the matrix B formed directly
when the numerical computations are
performed?

Look forward to your reply ...
--Rick Fierro

I am using Windows NT 4.0 and I want to install f2c. Can someone tell me:

(1) exactly which files I should download? (I know the web page and the readme file confused me.)

(2) How to install f2c?

Thanks.

Hi,

On my hobby time, I test the CPOLYDR (tompfs/419) which is supposed to give out the zeros of polynomial functions. I was surprised that dealing very simple one, x^2 + 2 x + 4=0, the results was worse. The printing out gives x1=1.7345 and its opposite. Has someones tried it before?

It looks as if Clement ran CPOLY in single precision. I ran it in double precision and
got the roots as -1 +/- sqrt(3) in all the
digits I printed out. It is not unusual to
lose at least 3 decimal digits of accuracy in
solving polynomial equations.

Versions of CPOLY and RPOLY in Fortran 90 style
can be found at my web site:
http://www.ozemail.com.au/milleraj

I have a large non-symmetric matrix eqn to solve. I read that conjugate gradient method requires symmetric positive definite matrix. Is there any variation of the method that I can use? Or any other numerical methods? Direct or iterative?

Hi,

Does anybody know the algorithms solving eigenproblem (faster than Jacoby- or QR -type) for complex matrices of a specific kind: A+i*B, where A is real symmetric, B is real diagonal and i is imaginary unit?
Is there a routine to handle the problem?

Thanks in advance.

I am interested in organizations that do research in Neural Networks. I am interested in specific institutes, companies, or organizsations that are doing research with sparce or limited data as well as applying neural networks to drilling applications. If someone knows of any organizations like this or if you are one could you please email me and send some specific research information.

I am interested in organizations that do research in Neural Networks. I am interested in specific institutes, companies, or organizations that are doing research with sparce or limited data as well as applying neural networks to drilling applications. If someone knows of any organizations like this or if you are one could you please email me and send some specific research information.

Ladies and Gentlemen:

The Rutishauser-Reinsch algorithm for partial
solution of the symmetric eigenvalue problem
originally published as ACM TOMS ALGORITHM 538
is currently available in Fortran 90/95.

Kindly see

<http://www.pjnikolai.homestead.com/index1.html>.

Paul J. Nikolai
Centerville, OH, USA

Hi,

I have the following data (rounded):
y x
28.0 1
31.5 2
36.0 3
42.0 4
50.4 5

I tried to fit a logistic function to the data by the least squares method but it did not work. I do not want to disclose the software I am using, because I do not really know what went wrong because the program just keeps going and going.

The logistic function I am using is:
Y = A/(B+exp(cx))

If anybody can do it, I would like to know how he or she did it and what kind of software he or she used. I am really curious to see what is going on.
Can anybody tell me what the result

Reply to Netvor - who does not seem to have left
an e-mail address.
The data show approximately exponential growth
with no suggestion of an upper asymptote. Thus
the value of c must be negative, somewhere near
c = -0.15. There is no information in the data
which will help you get sensible values for both
A & B. Just put B = 0, then you are fitting
Y = A.exp(-cx). Closer examination shows that
the growth is faster than exp(-cx). The growth
from x=1 to x=2 is by 12.5%, steadily increasing
to 20% between x=4 and x=5. The data don't fit
your model; why try to force it?

Alan Miller

Hi,
There is a linear programming problem that I am having difficulty with. Can you please help me.
It reads as follows:

The Sweet Tooth Candy Company manufactures and sells 2 products: Crunch Bars and Golden Goodie Bars. The company has a limited supply of 2 chocolate ingredients during the next month: 8,000 ounces of XJ100 and 7,000 ounces of XJ200. Each Crunch Bar uses 0.25 ounce of XJ100 and 0.5 ounce of XJ200. Each Golden Goodie Bar requires 0.5 ounc of XJ100 and 0.25 ounce of XJ200. In addition, the marketing department reports that while the company can seel all the Golden Goddie Bars it can produce, it can only sell up to a Maximum of 10,000 Crunch Bars. Sweet Tooth makes a profit of $0.05 on each Crunch Bar and $0.02 on each Golden Goodie Bar.

QUESTIONS:

1. You can buy more XJ100 and XJ200 from another supplier at a higher price ($0.05 extra per unce of XJ100 or XJ200). Will you purchase more XJ100 and XJ200 and how much? JUSTFY

2. The marketing department claims that a promotion campaign can be used to increse the maximum sale of crunch bars. How much are you willing to spend on such a campaign? JUSTIFY

3. If the production department has invented a new process to make Golden Goodie Bars that can cut production cost and thus increase profit margin of Golden Goodie Bars by $0.10 extra per bar, will that cut the original product mix? JUSTIFY

Hello,

I was wondering if anybody out there
had compiled the ARPACK library on a
Windows machine. If so, please let me know
which FORTRAN compiler you used and how
you compiled the library.

Thanks,
J-F. Seurin
Please cc to seurin@uiuc.edu

2001 WSES International Conference on:
Neural Networks and Applications (NNA '01)
http://www.worldses.org/wses/nna

- . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - .

2001 WSES International Conference on:
Fuzzy Sets & Fuzzy Systems (FSFS '01)
http://www.worldses.org/wses/fsfs

- . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - .

2001 WSES International Conference on:
Evolutionary Computations (EC '01)
http://www.worldses.org/wses/ec

- . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - . - .
Sponsored by: The World Scientific and Engineering Society (WSES)

The WORLDSES Conferences: NNA'01 (Neural Networks and Applications),
FSFS'01 (Fuzzy Sets Fuzzy Logic), EC'01 (Evolutionary
Computation) will take place at the Hotel TENERIFE PLAYA (main hotel for
the conferences) as well as at the Hotel SAN FELIPE. (Puerto De La Cruz,
Tenerife, Canary Islands (Spain), February 11-15, 2001).

Co-Sponsored by IIARD, IMCS


DEADLINE FOR PAPER SUBMISSION: OCTOBER 30, 2000
NOTIFICATION OF ACCEPTANCE/REJECTION: NOVEMBER 30, 2000