13010 Reputation

19 Badges

12 years, 295 days

MaplePrimes Activity

These are replies submitted by tomleslie

@Carl Love 

I fixed the typos and still got "inconsistent" results. I gave up on the OP's matrices,because they are just too big/ugly to deal with, so I constructed a simple example. I deliberately haven't loaded it as a worksheet because I'd like you to come up with the answers (true/false) to each of the following cases, before you cut/paste to Maple.

NB the test matrix is always symmetric.

M1:= < <0, x*(1+x)> | <0, 0> >:
M2:= < <0, 0> | <x+x**2, 0> >:
M1+M2; # whatever Maple says, this matrix is symmetric!
IsMatrixShape( M1+M2, symmetric);
IsMatrixShape( simplify(M1+M2), symmetric);
IsMatrixShape( simplify(M1)+simplify(M2), symmetric);
IsMatrixShape( simplify~(M1)+simplify~(M2), symmetric);
IsMatrixShape( simplify~(M1+M2), symmetric);
IsMatrixShape( expand(M1+M2), symmetric);
IsMatrixShape( expand(M1)+expand(M2), symmetric);
IsMatrixShape( expand~(M1)+expand~(M2), symmetric);
IsMatrixShape( expand~(M1+M2), symmetric)

How many did you get right??? - and can you explain why in each case??



No-one likes retyping or using cut/paste to figure out what you are trying to do. Please use the big green up-arrow (right-hand end of the second row in the toolbar) to upload the best worksheet you have so far

@Carl Love 

I got bogged down trying to fix this (cos I'm crap with 2D input). I ended up more or less converting/rewriting the OP worksheet in 1-D format, converting some indexed names to inert subscripts etc etc, and even then the final symmetric check would not run.

It does however provide the result false when using simplify, rather than map/expand??!!

See the last two execution groups on the attached

I used I__b in my worksheet, and have just changed my local version to M__b - I still can't get IsMatrixShape to admit that the matrix is symmetric unless I expand the entries

Check the simple example at the bottom of  the revised worksheet


My previous workheet uses matrices with symbolic entries, becuase I wanted to check the general case (and I still don't think that it should have been necessary to expand the entries within the product!)

However if your matrices contain floats (and these are a bit ugly), then I suppose that it is possible for round-off error in generating the entries of the product to result in a non-symmetric result.

See the attached worksheet.

It would seem that in order for IsMatrixShape(symmetric) to return true for an obviously symmetric matrix one first has to "expand" the entries in that matrix

I don't think this should be necessary - so I think you are right to be confused

Agree with everything Edgardo says:

  1. remove the curly braces from the equation definitions
  2. fix the unbalanced parenthesis in eqd3.
  3. supply the equations to the dsolve command as a list or set

I stuck a (more-or-less random) closing parenthesis in the definition of eqd3 in the attached worksheet, and fixed item (3) above, but Maple was then unable to find a solution for the four equations. I don't plan to try any harder until you supply synctactically correct equation definitions

Check the help entry for LinearAlgebra[GaussianElimination], which as the first sentence in its description says,

"The GaussianElimination(A) command performs Gaussian elimination on the Matrix A and returns the upper triangular factor U with the same dimensions as A."

Do you want ro reinvent the wheel for teaching/learning purposes?

@Carl Love 

Whilst agreeing Carl's post provides the correct answer, I thik the OP is correct to be curious. OP was using the defualt environment and I can't see why this doesn't work.

The third example in the attached sheet (which should be using the "default environment", gives a different answer from the first two examples - the "natural environment" or the "standard environment" and I can't figure out why.

It si particulary annoying becuase the first discrepancy occurs in the evaluation of M and things gets a bit pear-shaped from then on. However the eyeball evaluation of M in the third example suggests that it it is identical to that in the first two examples, just not simplified in the same way.

Something subtle going on???


At the risk of getting lunched by the serious mathematicians on here (and teaching you stuff you already know)- I think the following way:

If one has a second order inhomogeneous differential equation of the form


then the first textbook method to come up with a solution is to solve the corresponding homogeneous equation


then come  up with a "particular integral" - more or less any random function which is also a solution of the original equation. The complete solution is then the sum of the solution to the homogeneous equation and the "particular integral".

The second textbook method is to investigate the possibility of power series solutions, using something like Frobenius method. This gives rise to Bessel functions, Laguerre functions etc etc. Don't propose to discuss this here

With the first textbook method ie homogenous solution + particular integral approach, most textbooks will suggest techniques by which such a "particular integral" can be obtained. In textbook examples the form of the differential equation is usually such that the "particular integral" can be easily guessed, and the required integration easily performed - otherwise why put it in a textbook???. However the approach can be valid, even when the particular integral is "difficult",m and one gets a closed-form, ie non-power series solution

For this restricted set of second-order inhomogeous differential equations, solving the corresponding homogeneous equation is usually simpler than guessing the particular integral. Maple does  a pretty good job of coming with these solutions, inculding examples of particular integrals which I would never find. You just have to accept the following

  1. Most second order inhomogenous differential equations have no analytic solution known to man, Maple or anyone else (but can often be solved numerically)
  2. A (very) few second order inhomogenous differential equations have an analytic solution comprising the sum of the solution to the homogeneous equation and a particular integral
  3. Of these, even fewer have solutions where the particular integral can be obtained by average humans. Maple probably does a a fair job of getting the particular integral where most humans can't. (It is possible that Maple will fail to provide a particular integral when a seriously good mathematician could get one)
  4. A few more (but a long way from all)  second order inhomogenous differential equations which cannot be handled by the homogenous solution+particular integral approach, can be solved using Frobenius method to provide power series solutions
  5. Textbooks on second-order differential equations tend to leave the very misleading impression that second order inhomogeneous differential equations can *usually* be solved, because all the examples they give can be solved. Just remember that most can't be solved (analytically) by either method

The message to take away from this is don't get to hung up on the rather ugly integral term with an arbitrary variable like _Z in it. It looks unpleasant, but really it is not surprising and not anything to worry about. Think of it as a particular integral and be grateful that you got one.


I should apologise unreservedly for my original comments on this problem. By some unfortunate quirk of my browser your original post managed to fill the screen in such a way that

  1. the RHS was clipped (and my L/R scroll bar was off the bottom of the screen), so it didn't immediately occur to me to look for it - I just stupidly assumed that this was a posting problem
  2. to make things worse, the link to your worksheet happened to be off the bottom of the screen I was looking at, so I made the assumption that it didn't exist. How stupid that does make me?

I accept that getting the above wrong makes me a bit of an idiot.

The second issue about the clarity of your problem is, I think slightly (only slightly) more understandable. From your original problem statement, it wasn't obvious to me whether you were interested in

diff(rho(r),r)=piecewise( rho(r)<rho_crit, f(rho(r), m(r), K1, P1), f(rho(r), m(r), K2,P2))

diff(m(r),r)=piecewise( rho(r)<rho_crit, g(rho(r), m(r), K1, P1), g(rho(r), m(r), K2,P2))


diff(rho(r),r)= f(rho(r), m(r), K2,P2))

diff(m(r),r)=piecewise( rho(r)<rho_crit, g(rho(r), m(r), K1, P1), g(rho(r), m(r), K2,P2))

As far as I am concerned, either of these *could* have been a valid interpretation of the  problem you wanted to solve.

Before posting my response I gave the first of these a quick, dirty check and managed to get somethng which looked like a plausible solution. I then pondered why someone would be overlooking what seemed to be the obvious approach (piecewise) and be considering something really clever (like events, whihc I have never used), and concluded that maybe *I didn't really understand the issue* - hence my comment abot clarifying the problem. Although I was wrong on this as well. I think it is a forgiveable error - which my first criticism definitely wasn't

BTW I wouldn't generally recommend using piecewise definitions in functions which you supply to dsolve. The more-or-less inevitable dicontinuties in function slopes/values tends to make life very difficult for the solvers - but hey this time it worked!

PS Don't let my original rudeness put you off posting problems here. Most of the time, most of the people (including me) here are only too happy to help

I got a (probably justified) telling off earlier today for

  1. suggesting  that posters should use the big green up-arrow button on this entry page in order to upload code, and
  2. specifying more clearly what they are trying to achieve

I'm taking the risk of saying that both of these comments are justified in your case.

Having got that off my chest, I had a bit of play with your problem which you can see in the attached worksheet

It would seem that it it is relatively simple to generate real solutions, for x>-3. You may (or may not) be interested in what happens when x<-3, or x complex, I can't tell from your original post. However the attached sheet may give you some pointers with whatever you are trying to achieve


add( j, j in X*~evalf(map( ln, Y)));

If you want anyone to help, then I suggest that you upload your worksheet in such a way that

  1. it isn't clipped down the RHS
  2. Other posters don't have to do a lot of retying or cutting/pasting etc


You will also have to clarify your problem, because I (for one) don't really understand what you are trying to do.


For example, do you mean that your system is something like

diff(rho(r),r)=piecewise( rho(r)>rcrit, f(rho(r), m(r), K1, P1), f(rho(r), m(r), K2,P2))

diff(m(r),r)=piecewise( rho(r)>rcrit, g(rho(r), m(r), K1, P1), g(rho(r), m(r), K2,P2))

I'm not suggesting that you use the above, even if my interpretation is correct - just that you clarify exactly what it is that you want


@Carl Love OP asked for a plot of the function and "the parallelipiped approximation" on the same plot. I took this to mean a plot of the points in the matrix Z. Since the data is quite sparse 20x20, I found it unconvincing to *see* that the Z-data and function did in fact correspond, using a simple pointplot3d. I couldn't figure out how to interpolate a surface using a simple option with pointplot3d (this may be possible?), whereas such an option is more or less built-in to scatterplot3d

I accept that this approach falls into the category of "quick and dirty" and suspect that there is a better way. However I also thought this was a long way from being OP's biggest problem so decided not to worry about it

First 194 195 196 197 198 199 Page 196 of 199