For purposes of nearest neighbors this seems like an incredibly interesting shape to inscribe into:
The sphere, despite having spherical properties also maintains linear properties due to the corrugation. To me that means we can try to inscribe orthogonal properties into both of the spaces.
My understanding of these geometries isn't complex enough to make the connections, so my question is this:
Do you think its feasible to use shapes with this 'corrugated' property to make better nearest neighbor compression?
My intuition tells me that you can use the shape's linear nature to push apart independent components and inscribe the rest of the details into the spherical components. Or perhaps the opposite way.
I don't have any intelligent comments on your question, but I wanted to say that I am a fan of Quanta magazine, but somehow had missed this really cool article. So thanks for pointing me to this fascinating field. ;)
Good article, but it'll take some time to understand it. %1 is interesting, I used to use {..} for taking fractional part, %1 is intuitively easy, though not looking particularly good...
yeah. I think his website is extremely old and hasn’t been updated in the last decade or so. Despite this I linked to it because he is a legend in this field and so i think this is still the definitive reference.
As far as i understand, part of the story as to why dodecahedron and the cube fall short is due their non-triangular faces.
Can you explain the notation [0,1)^2 unit square, does the 2 represent the spatial dimensionality? So,[0,1)^3 is the unit cube? Why is 0 inclusive, but the 1 is exclusive?
"The first is that this mapping is area-preserving, not distance-preserving." Which area is being preserved?
Is there a volume preserving choice function?
What are points t0 and t3, are those the location of the singularity points? What is the definition of those "singularity points"? Is it that seeming void in the center of the fibonacci spiral? And that void doesn't exist within the unit square case?
So[0,1)^1 is a line interval, [0,1)^2 is a unit square and [0,1)^3 is the unit cube, and [0,1]^d is a d-dimensional cube.
2.Only one boundary can be included
It includes 0 but not 1 because it can only the context is usually that practitioners want a region where one edge will wrap to the opposite edge.
Thus they treat [0,1)^2 as if it is actually a 2-dimensional torus.
thus the the 2 boundaries acutally map to the same point, so you can only include one of them.
In our case as we are using x %1 = fractional part of x, the fractional part could be 0, if x=3.0, but it could never be exactly 1.
kevin948|4 years ago
I've had Nash's infinitely collapsible sphere stuck in my head for some time: https://www.quantamagazine.org/mathematicians-identify-thres...
For purposes of nearest neighbors this seems like an incredibly interesting shape to inscribe into: The sphere, despite having spherical properties also maintains linear properties due to the corrugation. To me that means we can try to inscribe orthogonal properties into both of the spaces.
My understanding of these geometries isn't complex enough to make the connections, so my question is this: Do you think its feasible to use shapes with this 'corrugated' property to make better nearest neighbor compression? My intuition tells me that you can use the shape's linear nature to push apart independent components and inscribe the rest of the details into the spherical components. Or perhaps the opposite way.
Hopefully that made sense!
extremelearning|4 years ago
avmich|4 years ago
Good article, but it'll take some time to understand it. %1 is interesting, I used to use {..} for taking fractional part, %1 is intuitively easy, though not looking particularly good...
extremelearning|4 years ago
Regarding the two-variable function mod(x,b). Typically this is written as x (mod b) in maths, and as x%b in computing.
It is generally well known that for positive integers x and b, the output of this function is the remainder when x is divided by b.
However, what is less well-known is that if b=1, then the convention is that:
x (mod 1) = x%1 = fractional part of x.
For example, Python, Excel both implement this special convention.
zarang|4 years ago
As far as i understand, part of the story as to why dodecahedron and the cube fall short is due their non-triangular faces.
sls|4 years ago
nmaleki|4 years ago
"The first is that this mapping is area-preserving, not distance-preserving." Which area is being preserved?
Is there a volume preserving choice function?
What are points t0 and t3, are those the location of the singularity points? What is the definition of those "singularity points"? Is it that seeming void in the center of the fibonacci spiral? And that void doesn't exist within the unit square case?
I especially enjoyed footnote #1.
extremelearning|4 years ago
So[0,1)^1 is a line interval, [0,1)^2 is a unit square and [0,1)^3 is the unit cube, and [0,1]^d is a d-dimensional cube.
2.Only one boundary can be included
It includes 0 but not 1 because it can only the context is usually that practitioners want a region where one edge will wrap to the opposite edge. Thus they treat [0,1)^2 as if it is actually a 2-dimensional torus.
thus the the 2 boundaries acutally map to the same point, so you can only include one of them. In our case as we are using x %1 = fractional part of x, the fractional part could be 0, if x=3.0, but it could never be exactly 1.
3) the mapping from the circle to the surface of the sphere is described here https://en.wikipedia.org/wiki/Lambert_azimuthal_equal-area_p...
the entire top edge of the square maps to the north pole, and the entire bottom edge maps to the south pole.
4.) t0 is the first point, t3 is the 4-th point.
Hope that helps!
nmaleki|4 years ago
zarang|4 years ago