MathGroup Archive 1999

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re(2): assumption, supposition?

  • To: mathgroup at smc.vnet.net
  • Subject: [mg15375] Re(2): [mg15331] assumption, supposition?
  • From: Andrzej Kozlowski <andrzej at tuins.ac.jp>
  • Date: Sat, 9 Jan 1999 23:58:28 -0500
  • Sender: owner-wri-mathgroup at wolfram.com

Pattern matching clearly is unsuitable for your purpose, since it
involves no deductions just mere "blind" substitution. However, one can
easily implement simple "deduction" of the kind you seem to want in
Mathematica, by making use of its ability to solve inequalities. There
are two p-ackages for this purpose, Algebra`InequalitySolve`  and
Algebra`AlgebraicInequalities`. However, the former one is limited only
to inequalities in one variable, so the latter one seems to be the
right one to use here. I have written some code which, seems to me, to
do the sort of thing that you might want. It is rather clumsy and very
imperfect as I spent only minutes on writing it. It's only meant to
suggest the kind of thing that can be done. I hope that some real
expert on these matters might be induced to comment, and perhaps even
to develop a real package that will properly do what I am only hinting
at. Anyway, here is the idea.
First we load the Inequality Solving package.

In[1]:=
<<Algebra`AlgebraicInequalities`

Next we define the function we shall use for testing conditions:

In[2]:=
cond[expr_List,var_]:=
  SemialgebraicComponents[Join[eepr,given],var]==
    SemialgebraicComponents[given,var]

This function has two parameters: a list of conditions (inequalities) to
be tested and a list of variables var. It depends on one global
variable, given, which is also a list of inequalities. What cond does
is to join the inequalities in expr to those in given and check if the
solution of the joint set is the same as that of given alone. If this
is the case than the conditions in expr forllow from those in given and
are thus satisifed.This is how it works. We first set a value for
given.


In[3]:=
given={x>0,y>1}
Out[3]=
{x>0,y>1}
Next we test if x+y>1 follows from "given". In[4]:=
cond[{x+y>1},{x,y}]
Out[4]=
True

On the other hand x+y>2 does not follow: In[5]:=
cond[{x+y>2},{x,y}]
Out[5]=
False

Next, we can define sqrt a replacement for Sqrt which will do what you
wanted. Really, of course, one should do this with Power, but I shan't
bother. Also, I introduce another global variable, variables, which
probably is  bad programming style, but again since this is only meant
as an illustration I shall not try to make it nicer:

In[4]:=
sqrt[x_^2/;cond[{x>0},variables]]:=x sqrt[x_]:=Sqrt[x]

Now we assign values to our global variables:

given={x>0,v>x,w>x+1};
variables={x,v,w};

We can see that sqrt behaves correctly on values which it knows as being
>0: In[5]:=
sqrt[w^2]
Out[5]=
w

On the other hand, in the case of variables it does not know anything
about it behaves as before:
In[6]:=
sqrt[u^2]
Out[6]=
      2
Sqrt[u ]


Andrzej



On Fri, Jan 8, 1999, Erk Jensen <Erk.Jensen at cern.ch> wrote:

>Probably I'm looking up the wrong keyword in Mathematica help, but I
>can't find what I'm looking for. Maybe that's because I've learned
>those expressions in german...
>
>My problem:
>
>I want Mathematica to assume that a certain condition is satisfied for
>the following algebraic transformations. More specifically: Let rho
>describe a radius coordinate. I know it is not negative real. So, for
>transformations of expressions containing rho, e.g.  Sqrt[rho^2], I
>want assure Mathematica that 0 <= rho is in fact satisfied, and that it
>is consequently allowed to replace Sqrt[rho^2] by rho. How do I do
>that?
>
>My solution was 
>
>Unprotect[Sqrt]; Sqrt[rho_^2] := rho/; 0 <= rho; Protect[Sqrt]; 
>
>rho /: 0 <= rho = True;
>
>and this seems to work, but I wonder whether this is really the proper
>way. Since the TagSet I'm using here assigns the whole statement to
>rho, and ?rho results in
>
>Global`rho
>rho /: 0 <= rho = True
>
>so I still have my doubts ...
>
>Can some of you experts enlighten me?
>
>Thanks in advance
>               -erk-

>Andrzej Kozlowski wrote:
>> 
>> It seems to me there is nothing wrong with what you are doing, but it is
>> rather limited. While indeed Sqrt[rho^2] will now return rho,
>> (rho^2)^(1/2) , (rho^3)^(1/3) will all behave as before. So you need to
>> modify the behaviour of Power rather than Sqrt. However, it seems to me
>> that the simplest way to achieve what you want is by using Abs[rho] in
>> place of rho in your formulas and then replacing Abs[rho] by rho in the
>> final output.  Mathematica knows that (Abs[rho]^2)^(1/2) is Abs[rho] etc.
>> 
>
>Thanks for your remark!
>1st part: it's even worse: if I set "rho /: 0 <= rho = True;" it will
>recognize
>0 <= rho as True, but not "rho >= 0" (!) so it doesn't actually do the
>checking, it just
>replaces "0 <= rho" by true.
>
>2nd part: this wouldn't help if my allowed range for rho would be say rho0 <=
>rho <= rho2!
>Which is actually what I have.
>
>I think I may go on with PowerExpand[Sqrt[x^2]] after all, that also
works for
>your
>examples ... but it's not exactly what I want. What I really want is best
>described by the
>option "Assumptions" of "Integrate"!
>
>Best regards,
>                 -erk-
>


Andrzej Kozlowski
Toyama International University
JAPAN
http://sigma.tuins.ac.jp/
http://eri2.tuins.ac.jp/



  • Prev by Date: Performance Problem
  • Next by Date: Re: Kalman filtering resources
  • Previous by thread: Re: Performance Problem
  • Next by thread: Re: Re(2): assumption, supposition?