How to pronounce“hyperrectangles”in English

Well, if there's N hyperrectangles, if I sum up all the surface areas of all of them, I should get 1.

And the claim is that if we want to approximate this with N hyperrectangles and we want to get a certain notion of the integral of the error,

First approximate the function with rectangles, then approximate it with hyperrectangles.

And then the final step is just to let this constant c grow really large so that we really do approximate the hyperrectangles.

And then I have the final term, which is the number of hyperrectangles I need.

This piece is the error per hyperrectangle.

Then show that a two-layer ReLU network can approximate a hyperrectangle.