1.7*10^308 is the max number supported by MATLAB. Any number exceeding this number is treated as infinity. Any method how to cope with this problem when you are expecting large numbers in your calculations.
I once found a nice package in R which addresses the problem of big number representation. It's in R, but the ideas are applicable to any language I guess.You can find the details here:
http://cran.r-project.org/web/packages/Brobdingnag/vignettes/brob.pdf
http://cran.r-project.org/web/packages/Brobdingnag/vignettes/brobpaper.pdf
Basically, you will need to make a custom representation of numbers and the limit will depend only on your imagination. For instance BB(111111) (Busy Beaver function) takes me 10 characters to write down (and not much in a code as well), yet it's insanely large. However, you will lose in accuracy. That's the price.
Read about Knuth's up-arrow notation as well.
From the given limit I deduce you are computing with floating point values (1.79...e308 is the maximum double value).
To keep computation time affordable and to avoid loosing accuracy, it might be a better approach to rescale your variables. If you're computing a physical model, it might also be a good idea to, e.g., introduce a unit-less representation of your primary variables.
Furthermore, keeping everything in the same range and avoiding too large numbers avoids other numerical problems.
I think using symbolic algebra computations allows you to use very large numbers instead of the floating point arithmetic programs.
I agree with Reuter. I found the remark that 10^308 is larger than the number of atoms in the Universe. What could be the sense in a larger number? Rescale! Otherwise you could also run in problems with condition numbers.
You can use "Variable Precision Integer Arithmetic" library in MATLAB. It can compute very big numbers. You can reach the library by the following link . It's easy to use. Just drag and drop into MATLAB then use it.
Ps: I suggest you MATHEMATICA for your computations. It's so speed and has great functional capabilities.
Good luck !
http://www.mathworks.com/matlabcentral/fileexchange/22725-variable-precision-integer-arithmetic
I agree with Cahit. I suggest you MATHEMATICA for your computations or use multiple precision toolbox in Matlab.
I need to work with billions of decimal digits. What program do you recommend? Does R-packages have that capacity?
- Badges
- Science topic
- Similar topics
- Computer Science
- Computing