Represent the line by the real numbers. By bijection, that's equivalent to [0,1], the closed unit interval. Now cut it into an unlimited (infinitely large) number of pieces. Name that number of pieces "H". Then [0,1] ~ Z* mod H where Z* is the hyperintegers and H is hyperfinite. This gives a useful intermediate object that captures things like a point at infinity in projective spaces (instead of nilsquare vectors, you can just designate some basis vectors infinitesimal length).

Let’s use it to link the discrete and continuous Fourier transform. The continuous is the standard part aka shadow of a hyperdiscrete Fourier transform so the roots of unity are sampled as i/H.

This is all brutally rough so feel free to request elaboration.

New to LessWrong?

New Comment
2 comments, sorted by Click to highlight new comments since: Today at 9:27 AM

Where can I read more about this perspective? 

I'm intrigued by the idea of linking the discrete and continuous Fourier transform through nonstandard analysis. 

The idea is certainly beautifully elegant - has it been worked out in more detail somewhere?