used both maple and mathematica for computer algebra, didn't find too much difference between them. Slightly different syntax, but you get over that fairly quickly

- actuaryalfred
**Posts:**144**Joined:**

The symbolic toolbox in Matlab is nice too. I always try to use it first before using Mathematica. I haven't used Maple so nothing could be said about that.For numerical stuff, Matlab is definitely better than Mathematica.

QuoteOriginally posted by: actuaryalfredFor numerical stuff, Matlab is definitely better than Mathematica.If you are talking about run-times could you post a few simple examples wherethe difference is dramatic? If you are talking about other issues, please elaborate.

Sir. Alan:Since I am a newbie on this forum, could you do me a favor for you to be a little more specific about the issue?I couldn't figure out what are you talking about!?TooNeat

My question made it sound like I had a specific issue in mind. I didn't. I was just asking actuaryalfred to elaborate on his remark. I know, as a Mathematica user, that Mathematica numerics seem faster these days than they were, say, 5 years ago (beyond clock speed improvements) and that WolframRI has worked on this.But as I am not a Matlab user, I can't compare them on my machines, and don't know ifcomparable speed improvements have been happening there.

Last edited by Alan on August 31st, 2006, 10:00 pm, edited 1 time in total.

I am a Maple user, it is the symbolic kernel of Matlab (but not thelatest version, so what ...). Generally it is satisfactory, but forsome integrals I have to consult MMA (where I am not sure, whetherresults are always true under any conditions [i.e. complex case],but it can be checked); reason seems to be: different approaches insome cases (too deep for a sketch).For numerics (usually 10 Digits in M, up to 14 Digits) most stuff ishandled through external libraries (NAG). One needs some experienceto get more out of it, but then it get really fast (forget anythingaround Bessel functions - a mess, hypergeometric numerics is to slowfor my taste). Newer versions allow to use the compiled version ofscripts (depends on coding style and complexity) on the fly.A really fine thing: one can look into the source code of almost allroutines, it is 'open'. And can generate code for VB, C, etc ...But not free of errors and I would not trust any of those CAS :-)A good source to see bugs and different opinions is sci.math.symbolic,sometimes it sounds like a religious war. In that newsgroup one willfind other system users as well and there are certainly some, whichare free and not bad at all. And sometimes good advice to the oldquestion "which one to choose?".

- actuaryalfred
**Posts:**144**Joined:**

QuoteOriginally posted by: AlanMy question made it sound like I had a specific issue in mind. I didn't. I was just asking actuaryalfred to elaborate on his remark. I know, as a Mathematica user, that Mathematica numerics seem faster these days than they were, say, 5 years ago (beyond clock speed improvements) and that WolframRI has worked on this.But as I am not a Matlab user, I can't compare them on my machines, and don't know ifcomparable speed improvements have been happening there.Couldn't come here for a couple of days. Seemed to be a shutdown for maintenance.When I said Matlab is better than Mathematica in the context of numerical computation I based the conclusion on my own experience and what I have heard. It may not be fair since I didn't use Mathematica for a long time and I didn't try writing programs on both and test the run-time. But I would say the way Matlab handles matrix should be better than Mathematica does. So by numerical calculation I should have said more specifically numerical calculation involving matrices. After all, Matlab and Mathematica should not be directly compared as they have been designed for different usage. Having said that, it's totally possible for one to catch up another. http://www.stats.uwo.ca/faculty/aim/epu ... lt.htmWhen I have time I'll also run those test programs in the above link in my laptop.

I am working with MatLab. Did a some theoretical stuff (partial analysis of some particular decay to obtain angular cross-sections) in Matematika for my PhD-thesis. Not big fun.Thinking in terms of integration into analitic work at the desk, I trust, MatLab is superior:- it gives me possibility to download Bloomberg data directly to MatLab,- I can make plots pretty quickly. Plotting capability is enourmous. It is somewhat worse of that I had in PAW (Physics Analysis Station) and ROOT used by physicists.- It is easy to prototype algortihms. With slight modifications of the code I can implement algo's into Java objects, then analyse/test it with MatLab again (to compare with prototype) and deploy it into standalone Java-application.Franckly, I would use ROOT, if bank would allow it.... as is it even more superior than MatLab and anything I can imagine...

- actuaryalfred
**Posts:**144**Joined:**

QuoteOriginally posted by: nikolFranckly, I would use ROOT, if bank would allow it.... as is it even more superior than MatLab and anything I can imagine...How about Python with scipy? In fact I'm now trying to use Python with scipy instead of Matlab in some situations. Found another interesting site:http://www.larssono.com/musings/matmatpy/index.html

QuoteOriginally posted by: actuaryalfredHow about Python with scipy?I am too old to learn the fashionable/new stuff. I need results and I need them quickly. ROOT is the best (actually, OO version of PAW, which covers all needed aspects of data analysis).CERNLIB is a free (sceintific) soft for all really practical things I ever encountered. NAG seems to sell same for money (I wonder how they managed to do that)...If I would have hedge fund under management, I would subscribe for CERNLIB and ROOT paying this guys for license. For two reasons:- support of science;- the argument that "professional" soft companies can serve me best does not work in critical situations, as they only suck money and one face problems anyway. CERNLIB code is the world longest validated scientific code by many participants...

QuoteOriginally posted by: actuaryalfredThe symbolic toolbox in Matlab is nice too. I always try to use it first before using Mathematica. I haven't used Maple so nothing could be said about that.For numerical stuff, Matlab is definitely better than Mathematica.If you use the Matlab symbolic toolbox, that is a stripped down version of Maple. Or it was the last time I checked.Personally, I find Matlab kind of kludgey. Many of the toolboxes appear to be like open source stuff without the peer review. Plotting is horrible after using plotting packages that have their shit together, like Igor (which is amazingly great, and almost completely unknown) or R. Still, the fact that it more or less works, and has been around forever ends up meaning a lot.Maple and Mathematica are sort of the same; computer algebra systems. Matlab is a wrapper program for EISPACK with add on modules.

- exneratunrisk
**Posts:**3559**Joined:**

My view: Computer algebra is only part of symbolic computation.Symbolic computation systems shall empower its users to manipulate (not only numbers, but) symbols, say, mathematical expressions, logical expressions and rules, geometrical objects, chemical objects, electrical curcuits, financial derivatives?,.... and even computer programs.Simplified, Numerical schemes transform functional symbols into numbers.Both are the core technologies for computer mathmatics.If this was true, Maple is "betweeen" computer algebra and symbolic computation.MatLab's symbolic toolbox is even closer to computer algebra, but MatLab benefits from a strong numerical kernel and a lot of (traditionally implemented?) toolboxes.Mathematica is THE representative of symbolic computation (or if you want the "language of mathematics). It has been extended with increasingly powerful numerical schemes in the last years.The good news, if you take their APIs and Link technologies you can combine them, combine them with the most advanced numerical foundations, say, NAG, and your proprietary solvers.When I think, an application can be described in the "language of mathematics" (functions and patterns), I choose Mathematica's declarative envoronment.

Last edited by exneratunrisk on November 22nd, 2006, 11:00 pm, edited 1 time in total.

For Quant Finance purposes though, such as visualizing the greeks of an exotic option that has no analytic solution available, how does Mathematica compares to Matlab in terms of:- speed?- learning curve for a non C++ programmer?

QuoteOriginally posted by: actuaryalfredThe symbolic toolbox in Matlab is nice too. I always try to use it first before using Mathematica. I haven't used Maple so nothing could be said about that.For numerical stuff, Matlab is definitely better than Mathematica.I also like to look first at Matlab's financial toolkit first, only to discover thatthe interface is not very friendly, so I end up using Mathematica every single time. I amcurrently seriously looking into avoiding the first look altogether.... It's become sort of a mantra to say 'For numerical stuff, Matlab is definitely better than Mathematica'.This probably was true about 10 years ago, in the pre-version 3 period of Mathematica. I am notconvinced that it is the case now. What seems to be true is that Matlab has accumulated a largerlibrary of available algorithms, but in principle, I am not sure if 'anything you can do I can do at least as good'doesn't apply here.BR,yossi

GZIP: On