Re:Why do roots of polynomials tend to have absolute value 1?
by dnautics on Sunday October 05, 2014 @12:44PM
(#8412691)
A key point is that the distribution used to generate the coefficients is the uniform distribution over an interval. What happens if, say, a cauchy distribution (centered around 0) were used?
[ Parent
| on HN ]
Re:Why do roots of polynomials tend to have absolute value 1?
by archgoon on Sunday October 05, 2014 @01:09PM
(#8412753)
randomPoly[n_, x_, {a_, b_}] :=
x^Range[0, n].RandomVariate[CauchyDistribution[2, 1], n + 1];
Show[Graphics[
Point[{Re[x], Im[x]}] /. NSolve[randomPoly[300, x, {27, 42}], x],
Axes -> True]]
http://imgur.com/a/52oEs#0The unit circle is still present, but it now seems to be getting a few uniform outliers, a sort of halo.
I think that the explanations given in the responses are not correct. Even for the unit circle, they account for why you don't seem many points outside of the interval, but fail to explain why you rarely see points inside of the unit circle. The argument essentially is that inside the unit circle the equation reduces to
a1*x+a0 = 0
But the distribution for this (randomly picking a1 and a0 when you generate the poly) indicates you'd expect the occasional circle that had a point that was about .3 away from the radius. If you generated the circle 10 times, you don't. You rarely get as far as 0.1 .
Re:Why do roots of polynomials tend to have absolute value 1?
by dnautics on Sunday October 05, 2014 @08:13PM
(#8413864)
is it also not weird that the outliers are approximately the same radius out from each other - and also seem to be evenly spaced around the circle??
Re:Why do roots of polynomials tend to have absolute value 1?
by stephencanon on Sunday October 05, 2014 @02:19PM
(#8412943)
Steve Zelditch has a nice talk exploring some more reasonable distributions: http://www.math.northwestern.edu/~zelditch/Talks/AIM.pdf