approximation

Approximation is the art of swapping a hard object for an easy one that is close enough — like sketching a coastline with straight line segments. The picture is wrong in detail, but if the error is controlled, conclusions about the simple version transfer to the real thing.

In analysis, approximation means replacing an object (a number, a function, a curve) by a simpler one within an explicitly controlled error, and then letting the error shrink to zero. This is the deep pattern beneath the whole subject: a derivative approximates a function by its tangent line, a Taylor polynomial approximates a smooth function near a point, an integral is approximated by Riemann sums, and the reals themselves are approximations (limits) of rationals.

The decisive word is ‘controlled’. An approximation is worthless unless paired with an estimate of how far off it is; the power comes from making that error arbitrarily small. The honest caveat: an approximation that converges pointwise can still fail to preserve properties of the target — pointwise limits of continuous functions need not be continuous — so the type of control (pointwise versus uniform) matters as much as the fact of convergence.