Re: Symbolic solve in Octave isn't the same as in MATLAB

[Colin Macdonald]

On 02/04/17 04:44 PM, spertasolutions wrote:
> I'm using the following OCTAVE command to solve an equation:
>
> syms lambda
> f = 0.9999999999999999614356461247534*lambda^8 +
> 0.010136594318889089056437488043559*lambda^7 +
> 0.000015602822908760063058460837969499*lambda^6 -
> 0.000000047482181730798751628339904656436*lambda^5 -
> 1.6804583575154387080270563605856e-10*lambda^4 -
> 1.9779236089667689638174134679895e-13*lambda^3 -
> 1.1091457134190232245774382079799e-16*lambda^2 -
> 2.9823717782777009154185183839051e-20*lambda -
> 3.202469692780878591525091765936e-24
> solve(f==0, lambda)
>
> The answer I get is
>
> ans = {}(0x0)
> On the other hand, when I try to use the same command in MATLAB, I get the
> following answer:
>
> ans =
>  0.0027518081408059183214532684717853
> -0.0075831287286429577224408168637228
> - 0.00029575583281255452382838468617809 -
> 0.0001257455668418381216048479457788*i
> - 0.0016381334730465236707629472401125 -
> 0.00001988265276086766636376330275894*i
> - 0.0016381334730465236707629472401125 +
> 0.00001988265276086766636376330275894*i
> - 0.00071874755966694673136475169368249 -
> 0.00019245511322083820223381924903732*i
> - 0.00029575583281255452382838468617809 +
> 0.0001257455668418381216048479457788*i
> - 0.00071874755966694673136475169368249 +
> 0.00019245511322083820223381924903732*i
> Is there any way I can make OCTAVE's output equal to MATLAB's output, may be
> with some tweaks?

Not that I know of.  There is no formula for roots of a 8th degree 
polynomial.  I'd suggest using "vpasolve".  Perhaps this is what 
Matlab's Symbolic Toolbox is doing anyway.

Some other comments:

1.  You're mixing floating point and symbolic quantities in the same 
expression: in my experience, that's probably a bad idea.

2.  For example: the roots of that 8th degree polynomial are going to be 
very sensitive to error in the monomial coefficients, whether solved for 
numerically or symbolically.

best,
Colin