Question: Exploring Symbolic Computation with Maple: From Exact Expressions to Numerical Insight

One of the things I find particularly interesting about Maple is the way it allows us to move between symbolic and numerical mathematics without treating them as completely separate worlds.

Consider a relatively simple problem such as

restart:

f := x -> exp(-x^2)*sin(x):

df := diff(f(x), x):
d2f := diff(f(x), x$2):

simplify(df);
simplify(d2f);
The symbolic expressions obtained from these calculations are exact. This is useful not only because Maple can perform the differentiation automatically, but also because the resulting expressions can be manipulated further before any numerical approximation is introduced.

For example, we can investigate the stationary points of the function by solving

solve(df = 0, x);
or, when an exact solution is not convenient, use numerical methods:

fsolve(df = 0, x);
This distinction between solve and fsolve is quite important in practical symbolic computation. The first attempts to preserve the mathematical structure of the problem, while the second focuses on obtaining numerical roots.

We can also examine the same function graphically:

plot(f(x), x = -4 .. 4,
     labels = ["x", "f(x)"]);
The interesting part is that the graphical result is only one representation of the problem. We still have access to the exact derivative, second derivative, algebraic transformations, limits, series expansions, and numerical evaluations.

For instance, the Taylor expansion around zero can be obtained with

series(f(x), x = 0, 8);
which gives another way of understanding the local behaviour of the function.

This leads to a useful workflow in Maple:

symbolic expression → exact transformation → analytical investigation → numerical evaluation → visualization

Rather than replacing mathematical reasoning, computer algebra can make it easier to move between these different representations.

I also find this approach useful for more complicated problems. Once the symbolic structure has been obtained, individual parts of the expression can be simplified, approximated, plotted, or evaluated independently. This can make it easier to identify where a numerical result comes from instead of treating the numerical output as a black box.

For me, this is one of the most useful aspects of Maple: the same mathematical object can be explored from several perspectives while keeping the exact form available whenever it is useful.

I would be interested to hear how other Maple users approach the symbolic-to-numerical transition, especially for larger expressions where a fully symbolic solution becomes impractical.