@J.M. @J.M. It just seems darn lucky for humans that geometry implies $\cos(n\,\Arctan x)$ is a polynomial, which leads to Fourier series that converge exponentially, and then FFT and aliasing make it computationally efficient. I actually woke up this morning thinking this. What fun. :)

@Michael, it's all tightly interwoven math, so it's not supposed to be surprising. Still, I can't help but feel that there's a bit of magic in at least that corner of math… :)

@J.M. I am trying to find all points where a 2D (or 3D) scalar function has zero gradient (within a rectangle). I need to make it work relatively fast (Manipulate) speed. What are my options? There's this, i.e. use ContourPlot to find starting points. And there's Reduce, but Reduce will be too slow. So is a ContourPlot based solution my best bet?

@Szabolcs If you don't have good starting points, getting ContourPlot[] (with MeshFunctions) to guess for you is as good a strategy as any.

(For the 3D case, there's this.)

Thanks!

@MarianoSuárez-Alvarez You mean "real-world" network? No, GraphData is for small graphs and their mathematical properties. (BTW it got a big update in the soon to be available version 11.) For real-world networks, check ExampleData["NetworkGraph"] and the many network repositories online.

networkdata.ics.uci.edu/index.php snap.stanford.edu/data nexus.igraph.org www-personal.umich.edu/~mejn/netdata konect.uni-koblenz.de and many others