3 ms·
I'm pretty good with math but this article is making absolutely no sense at all to me. I'd like to focus mainly on just the volume-of-a-sphere and corner-of-cub
by crazygringo 4y ago
I'm pretty good with math but this article is making absolutely no sense at all to me. I'd like to focus mainly on just the volume-of-a-sphere and corner-of-cubes aspects if someone here can clarify?
Because the idea is that the "volume of a d-sphere goes to 0 as d grows". The chart as provided seems nonsensical -- because the unit of volume is different in every dimension. What does it even mean to say to compare the area of a circle with the volume of a sphere? Because they don't have the same units.
The best I can guess is that it's attempting to plot the volume of a d-sphere of radius r as a proportion of the volume of a d-cube it fits inside of with edge length 2r, so it is now unitless. Is that correct?
So in that case, a circle with radius 1 occupies 79% of its square, and a sphere occupies 52% of its cube, and a 3-sphere occupies 31% of its hypercube. All of which is fairly intuitive, that the sphere takes up proportionally less space of a hypercube the more dimensions you add. This is what you'd expect, because we already see it from 2 to 3 dimensions. And the intuition here is easy, because adding each dimension is an "extrude-and-slice" operation -- extrude the current shape along the new dimension (which preserves proportion), then slice away with a freely spinning circle-cutter (which decreases proportion).
But for some reason the chart shows sphere volume increasing up to 6 dimensions before it decreases, which I don't understand at all.
And then it attempts to show this "graphically" by drawing a kind of concave rhombus figure, which is even more mystifying, because every point on a hyphersphere is still the same distance from the center. There are no "corners"!
And that figure is next to a drawing of a hypercube that looks more like a star with the explanation "Notice that the corners of the cube are much further away from the center than are the sides." But this is also given zero explanation whatsoever, and a hypercube is the easiest thing in the world to visualize in higher dimensions because a unit hypercube simply stretches from 0 to 1 in every dimension.
I fail to see how the corners are "much farther away". According to what measure? Because the distance from the center of a unit square to its corner is .707 units. The distance from the center of a unit cube to its corner is .866. The distance in 20 dimensions is 2.236, and in 100 dimensions is 5.0. So yeah the corners get a little bit farther away -- exactly as you'd expect intuitively -- but the increase is also extremely slow, as you'd expect. And a distance of 5.0 in a hundred dimensions doesn't really seem "much" farther. It's a mere factor of 7 after adding a whole 98 additional dimensions. If anything, this shows that corners always stay "close by" no matter how many dimensions you add.
So I'm utterly failing to see what's supposedly so counterintuitive about high dimensional spaces at all here. This article baffles me completely.