>> And of course many useful transformations will not be found anyway.
>> For instance, I don't know of any practical way to perform
>> "simplifications" like x^5-14*x^4+94*x^3-264*x^2+409*x-242 =>
>> (x+1)^4+(x-3)^5.
>>
>
>
This might not be quit so hard, if you try some heuristics that are
modeled on "completing the square".
It is certainly worth finding such rearrangements in the context of
"single use expressions" because the second arrangement
will generally have much tighter error bounds if x is an interval. (It
is the sum of 2 single-use expressions....)
RJF