headers
Fabrizio Caruso | 1 Feb 2005 16:31
Picon
Picon

fast_linsolve

Dear Maxima Mailing List

The in line manual shows an entry for 
"fast_linsolve" which is (or was) an optimized 
linear system solver for sparse matrices.

For me it would be great to have it
because my summation package would
have a considerable speed-up.
I am sure many other packages would 
take advantage of it.

Unfortunately "fast_linsolve" seems
to have disappeared from Maxima.
Does someone know what happened to it?
Is it still there under a different name?
Anyone working on sparse linear systems
(where entries can possibly be rational functions)?

    Regards

    Fabrizio
Permalink | Reply | 6 comments |
headers
David Ronis | 1 Feb 2005 18:52
Picon

Bug in today's CVS


I'm trying to use maxima to do a saddle point integration with the following:

f(x):=x-2*log((exp(x)+1)/2) + x**2/4;
exp(-x**2/4)*taylor(exp(f(x))*(x*kb+mu)**gamma,x,0,4);
integrate(%,x,%minf,%inf)/4;

As far as I can tell, the taylor expansion is done correctly, however,
the multiplication by the additional exponential in the second line
seems to be ignored.  The resulting integral doesn't converge and I'm
left with explicit %inf/%minf's at the integration (which is wrong in
any event).

David
Permalink | Reply | 2 comments |
headers
Barton Willis | 1 Feb 2005 19:14
Favicon

Re: Bug in today's CVS

Taylor expansions are contagious.  When maxima multiplies exp(-x^2/4) 
times a
Taylor polynomial, the product gets expanded as a Taylor polynomial. To
do the calculation, use 'ratsimp' to convert from a Taylor polynomial to 
a non-taylor.

Also, I think you want 'minf and 'inf instead of %minf and %inf.

(%i3) kill(all)$
(%i1) f(x):=x-2*log((exp(x)+1)/2) + x**2/4$
(%i2) exp(-x**2/4)*ratsimp(taylor(exp(f(x))*(x*kb+mu)**gamma,x,0,4))$
(%i3) display2d : false$
(%i4) integrate(%o2,x,'minf,'inf)/4;
Is  gamma+4  positive, negative, or zero?

pos;Is  mu  positive or negative?
pos;Is  gamma  an integer?
no;Is  gamma+3  positive, negative, or zero?
pos;Is  kb*gamma  positive, negative, or zero?
pos;Is  gamma+2  positive, negative, or zero?
pos;Is  kb^2*gamma  positive, negative, or zero?
pos;Is  gamma+1  positive, negative, or zero?
pos;Is  gamma  positive, negative, or zero?
pos;Is  kb  zero or nonzero?
nonzero;Is  kb  positive or negative?

pos;(%o4) (24*sqrt(%pi)
         *(8*kb^4*mu^gamma*gamma^4-48*kb^4*mu^gamma*gamma^3
                                  +88*kb^4*mu^gamma*gamma^2
                                  -48*kb^4*mu^gamma*gamma+2*mu^(gamma+4))
(Continue reading)

Permalink | Reply | 0 comments |
headers
Richard Fateman | 1 Feb 2005 19:29
Picon
Favicon

Re: Bug in today's CVS

What is %minf?  Do you mean minf?  Same for inf.

Usually people use x^y rather than x**y.

When you take a taylor series Q(x) and multiply it
by exp(x),  the exp(x) is converted to a taylor series
and multiplied.
Assuming you do not want this,  you might try
something like

expand (exp(-x^2/4)* ratdisrep(taylor ...)))

integrate(%,x,minf,inf)

David Ronis wrote:

> I'm trying to use maxima to do a saddle point integration with the following:
> 
> f(x):=x-2*log((exp(x)+1)/2) + x**2/4;
> exp(-x**2/4)*taylor(exp(f(x))*(x*kb+mu)**gamma,x,0,4);
> integrate(%,x,%minf,%inf)/4;
> 
> 
> As far as I can tell, the taylor expansion is done correctly, however,
> the multiplication by the additional exponential in the second line
> seems to be ignored.  The resulting integral doesn't converge and I'm
> left with explicit %inf/%minf's at the integration (which is wrong in
> any event).
> 
> David
(Continue reading)

Permalink | Reply | 1 comment |
headers
Wolfgang Jenkner | 1 Feb 2005 18:47
Picon

Re: fast_linsolve

Fabrizio Caruso <fabrizio.caruso <at> math.univ-rennes1.fr> writes:

> Dear Maxima Mailing List
>
> The in line manual shows an entry for "fast_linsolve" which is (or
> was) an optimized linear system solver for sparse matrices.

> Unfortunately "fast_linsolve" seems to have disappeared from Maxima.
> Does someone know what happened to it?  Is it still there under a
> different name?

It's still there in share/affine (written by Bill Schelter).  The CVS
version is currently broken (mainly because of the sloop -> loop
change).  The version contained in the last stable release is supposed
to contain some unspecified but reasonable amount of working stuff.

If you use the stable version try `load("affine.lisp");'.  Probably,
this won't work because of some brain-dead logical-pathname stuff of
mine.  Plan B: Batch a file containing the following

load("affine/compat");
load("affine/amacros");
load("affine/dummy-gc");
load("affine/polybas");
load("affine/sparsemat");
load("affine/aquotient");
load("affine/new-rat");
load("affine/polya");
load("affine/ndotsimp");
load("affine/polyb");
(Continue reading)

Permalink | Reply | 5 comments |
headers
Richard Fateman | 1 Feb 2005 22:24
Picon
Favicon

macsyma on sale on ebay!

http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=3789&item=7130355446&rd=1&ssPageName=WDVW

$500.  I just came across this and thought you guys might be
interested.

Don't rush, there are 99 copies available.

RJF
Permalink | Reply | 2 comments |
headers
Paulo Ney de Souza | 1 Feb 2005 22:46
Picon
Favicon

Re: macsyma on sale on ebay!


What a bargain!

Paulo Ney

    >From maxima-admin <at> math.utexas.edu  Tue Feb  1 13:25:32 2005
    >From: Richard Fateman <fateman <at> cs.berkeley.edu>
    >To: maxima <maxima <at> math.utexas.edu>
    >Subject: [Maxima] macsyma on sale on ebay!
    >
    >http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=3789&item=7130355446&rd=1&ssPageName=WDVW
    >
    >$500.  I just came across this and thought you guys might be
    >interested.
    >
    >Don't rush, there are 99 copies available.
    >
    >RJF
    >
    >_______________________________________________
    >Maxima mailing list
    >Maxima <at> www.math.utexas.edu
    >http://www.math.utexas.edu/mailman/listinfo/maxima
    >
Permalink | Reply | 0 comments |
headers
C Y | 1 Feb 2005 22:59
Picon
Favicon

Re: macsyma on sale on ebay!

Heh - I guess David Schmidt is 71271.1417 on ebay.  Well, it's the
first active marketing I've seen from them since 1999, so that's gotta
be a good sign.  99 copies available after almost eight days...  I
guess there isn't a rush to buy $500 software with no hard copy
documentation!

Hmm, new spiffy website for http://www.symbolics.com too.  Looks like
anybody wanting to upgrade to 2.4 from an earlier version can do it for
$250.

I hope they start active development of Macsyma again - it would really
be a shame to have it remain static indefinitely.  Maybe they could at
least started digging out the Linux version from whatever server it's
sitting on and help some of those folks that can't use their older
purchased copies. 

CY

--- Richard Fateman <fateman <at> cs.berkeley.edu> wrote:

>
http://cgi.ebay.com/ws/eBayISAPI.dll?ViewItem&category=3789&item=7130355446&rd=1&ssPageName=WDVW
> 
> $500.  I just came across this and thought you guys might be
> interested.
> 
> Don't rush, there are 99 copies available.
> 
> RJF
>
(Continue reading)

Permalink | Reply | 1 comment |
headers
Jim Schott | 2 Feb 2005 15:21
Picon

Simplification into a specific form


I have a rather lengthy function of several variables that I know simplifies
into a particular form. An over-simplified version of the problem goes like
this. Suppose I know a function of x and y simplifies into the form

a*(x + y)^2 + b*(x^2 + y^2)

where a and b are constants. If maxima has the function in the form

2*(x^2 + y^2 + x*y)

how can I get it to give me it in the desired form? That is, how do I
get maxima to tell me that a=1 and b=1?

Any suggestions would be appreciated.

Jim Schott
Permalink | Reply | 3 comments |
headers
Richard Fateman | 2 Feb 2005 17:46
Picon
Favicon

Re: Simplification into a specific form

Jim Schott wrote:

> 
> I have a rather lengthy function of several variables that I know 
> simplifies
> into a particular form. An over-simplified version of the problem goes like
> this. Suppose I know a function of x and y simplifies into the form
> 
> a*(x + y)^2 + b*(x^2 + y^2)
> 
> where a and b are constants. If maxima has the function in the form
> 
> 2*(x^2 + y^2 + x*y)
> 
> how can I get it to give me it in the desired form? That is, how do I
> get maxima to tell me that a=1 and b=1?
> 
> Any suggestions would be appreciated.
> 
> Jim Schott
> 
> _______________________________________________
> Maxima mailing list
> Maxima <at> www.math.utexas.edu
> http://www.math.utexas.edu/mailman/listinfo/maxima

Let k1=x^2+y^2
     k2=x+y

Then try   ratsubst(k1,x^2+y^2,Expression);
(Continue reading)

Permalink | Reply | 1 comment |

Gmane