On 4/25/2012 10:08 AM, Henry Baker wrote:
> I was curious about the current state of the problem of simplifying constant expressions.
>
> I realize that the problem is unsolvable, in general, but I wonder if there are classes of solutions that exist.
You can find work by Daniel Richardson and others involving Schanuel's
conjecture and log-exp expressions
with a google search.
>
> Basically, I'm talking about finite expressions built up from constant integers, using +,-,*,/,exp(x),log(x),expt(x,y), sqrt(x), etc. Ditto for additional standard constants like %pi, %e, %i, etc.
>
> I'm not talking about floating point numbers here, but exact computations.
>
> I'm not talking about "for all" or "there exists" types of computations, but merely constant expressions.
There's also work that can be done by evaluating expressions as floating
point numbers and then trying
to find another expression that evaluates to the same number, see for
example,
http://www.cecm.sfu.ca/organics/papers/bailey/
>