Another question that has been bugging me for a while:
Consider a vector space of all piecewise continuous functions (more specifically, those that satisfy the Dirichlet conditions) defined on an interval of the reals, for example f:(-pi, pi) => R.
This space has (I think!) a number of dimensions equal to the cardinality of the continuum, alef[1].
However, consider the Fourier series approximation:
g(x) = a[0] + sum(n = 1, 2, ...)(a[n]*cos(nx) + b[n]*sin(nx))
There are a countable number of parameters for g, namely: a[0], a[1], ..., and b[1], b[2], ...; in other words, the space of all such g's has aleph[0] dimensions.
How is it possible that for every such f, there is a corresponding choice of the a's and b's?
Treated a slightly different way: There are alef[2] such f's, and there are only alef[1]^alef[0] = alef[1].
I asked one of my prof's, and he pointed out that although the set of rationals is smaller than the set of reals, any real number can be approximated arbitrarily close by the rationals. Using that analogy, he argued that any f(x) can be approximated arbitrarily close by a g(x). However, there is a ton of hand-waving here, and I'm not sure whether I'm convinced.
Another resolution to this issue would be if there are only countably many dimensions for the vector space of all f's satisfying the Dirichlet conditions. However, I think that, at least in the case where the domain for f is the whole number line, there are uncountably many dimensions; please correct me if that is my error.