Re: Root Finder vii.
From: Jon G. (jon8338_at_peoplepc.com)
Date: 11/17/04
- Next message: Han de Bruijn: "Re: Cantor's diagonal proof wrong?"
- Previous message: Chris Menzel: "Re: Cantor's diagonal proof wrong?"
- In reply to: Jon G.: "Root Finder vii."
- Next in thread: C. Bond: "Re: Root Finder vii."
- Reply: C. Bond: "Re: Root Finder vii."
- Messages sorted by: [ date ] [ thread ]
Date: Wed, 17 Nov 2004 12:09:59 GMT
That is,
The polynomial
a[0]+a[1]t+a[2]t^2+a[3]t^3+ ... + a[n] = 0
may be expressed as the dot product of the two vectors,
T = (t,t^2,t^3,...,t^n) and
N = (a[1],a[2],a[3],...,a[n]) or,
(t,t^2,t^3,..,t^n)*(a[1],a[2],a[3],..,a[n])+a[0]=0 or
T*N+a[0]=0 is the nth degree polynomial
N may be considered the normal to the more general case
of the plane,
a[0] + a[1]x[1] + a[2]x[2] + ... + a[n]x[n] = 0
where x[1]=t x[2]=t^2 x[3]=t^3 ... x[n]=t^n
Following T up from the origin, R is the variable vector
parallel to N and orthogonal to T-R, as T varies along
its curve. then
R*(T-R)=0
taking the derivative with respect to t,
R'*(T-R)+R*(T'-R')=0
or
2R'*R = R'*T + R*T'
The projection of T' onto N/|N| is R'
T'*N N
---- ---- = R'
|N| |N|
define
Q = (T*N)N/|N|^2 = -a[0]N/|N|^2
Q = (-a[0]/|N|^2)(a[1],a[2].a[3],..a[n])
or
Q = (q[1],q[2],q[3],...,q[4])
Since R' is in the same direction as Q, it holds that
since Q and T'-(T'*N)N/|N|^2 are orthogonal,
Q*(T'-(T'*N)N/|N|^2) = 0
where Q is the shortest vector from the origin to the plane.
T = (t,t^2,t^3,...,t^n)
T'= (1,2t,3t^2,..,nt^(n-1) )
T'*Q =q[1]+2q[2]t+3q[3]t^2+....+ nq[n]t^(n-1)
=(1/t)T*(q[1],2q[2],3q[3],...,nq[n])
T'*N = a[1]+2a[2]t+3a[3]t^2+....+ na[n]t^(n-1)
=(1/t)T*(a[1],2a[2],3a[3],...,na[n])
Suppose the vectors D,S,U are defined,
D=( 1 , 2 , 3 ,...,n )
S=(a[1],2a[2],3a[3],...,na[n])
U=( 1 , 1 , 1 ,...,1 )
Then
T'* U =(1/t)T*D
T'* N =(1/t)T*S
for instance,
(1,2t,3t^2,..,nt^(n-1) )*(1,1,1,..)
=(1/t)(t,t^2,t^3,...,t^n)*(1,2,3,..)
and
(1,2t,3t^2,..,nt^(n-1) )*(a[1],a[2],a[3],..)
=(1/t)(t,t^2,t^3,...,t^n)*(a[1],2a[2],3a[3],..)
dividing the below equations with each other,
T'* U =(1/t)T*D
T'* N =(1/t)T*S
(T'*U)(T*S) - (T'*N)(T*D) = 0
factoring out T,
T*((T'*U)S - (T'*N)D) = 0
so T is orthogonal with (T'*U)S - (T'*N)D
or T is orthogonal with (1/t)((T*D)S - (T*S)D)
or T is orthogonal with (T*D)S - (T*S)D
this is true, since
T*((T*D)S - (T*S)D) = 0
Let,
G = (T*D)S - (T*S)D
If m is some ratio of G such that,
(mG-Q)*N = 0
mG*N - Q*N = 0 since Q*N=T*N,
m[(T*D)(S*N)-(T*S)(D*N)]-(T*N) = 0
or
T*[m((S*N)D-(D*N)S )-N] = 0
One solution for m, among others, can be found from
m((S*N)D-(D*N)S - N = 0 or
m((S*N)D-(D*N)S = N taking the dot product of both
sides with D and solving for m,
N*D
m = ---------------------
(S*N)(D*D)-(D*N)(S*D)
Otherwise, T and m((S*N)D-(D*N)S )-N are orthogonal
Notice that since N*((S*N)D-(D*N)S )= 0,
((S*N)D-(D*N)S ) is orthogonal to N.
T is orthogonal with m((S*N)D-(D*N)S)-N
N is orthogonal with (S*N)D-(D*N)S
OY=m((S*N)D-(D*N)S)-N
Y
*------------ N T
\ | /
\ | /
\ | /
\ | /
\ | /
\ | /
Z _________\|/
|O
OZ= |
(S*N)D-(D*N)S |
|
|
|
-N
T is orthogonal with OY
N is orthogonal with OZ
since angle YOZ = angle TON,
OY*OZ T*N OY*N
-------- = ------ = {1 - (-------)^2 }^(1/2)
|OY||OZ| |T||N| |OY||N|
(OY*OZ)^2 (OY*N)^2
------------ = 1 - -----------
|OY|^2|OZ|^2 |OY|^2|N|^2
Multiplying both sides of this equation by
|OY|^2|OZ|^2|N|^2,
(OY*OZ)^2|N|^2 = |OY|^2|OZ|^2|N|^2 - (OY*N)^2|OZ|^2 eqn i.
Letting (S*N)D-(D*N)S = C, OZ=C and OY=mC-N and
(OY*OZ)^2=((mC-N)*C)^2|N|^2 = m^2|C|^2|N|^2
|OY|^2=(mC-N)*(mC-N) = m^2|C|^2+|N|^2
|OZ|^2=|C|^2
(OY*N)^2=((mC-N)*N)^2=-|N|^4
Substituting these in eqn i,
(m^2|C|^2|N|^2)|N|^2
=
(m^2|C|^2+|N|^2)|C|^2|N|^2
+
|N|^4|C|^2 or
m^2|N|^2 = m^2(|C|^2+|N|^2) + N^2
N^2 = m^2(|N|^2 - |C|^2 - |N|^2)
|N|
m = - -----
|C|
So
OY=m((S*N)D-(D*N)S)-N or
|N|
OY= - ---((S*N)D-(D*N)S)-N
|C|
or
|N|
OY= - ---------------((S*N)D-(D*N)S) - N
|(S*N)D-(D*N)S|
or
|N|
OY= - --- C - N
|C|
(qOY-Q)*N=0
qOY*N-Q*N=0
q(0-|N|^2) = Q*N
q(-|N|^2) = -a[0]
q = a[0]/|N|^2
angle ZOY
OY*OZ T*N
cos(ZOY)=-------- = ------ = cos(NOT)
|OY||OZ| |T||N|
T*N= -a[0] and
a[0]|OY||OZ|
|T| = - -------------
|N|(OY*OZ)
(a[0])^2|OY|^2|OZ|^2
|T|^2 = ---------------------
|N|^2(OY*OZ)^2
|OY|^2 = 2|N|^2
(OY*OZ)^2 = (|N||C|)^2
(a[0])^2 2|N|^2 |C|^2
|T|^2 = --------------------- = 2|Q|^2
|N|^2 |N|^2 |C|^2
T = Q + {|T|^2-|Q|^2}^(1/2)(-C/|C|)
|Q|
T = Q +/- ----- C
|C|
Where
C = (S*N)D-(D*N)S
D=( 1 , 2 , 3 ,...,n )
S=(a[1],2a[2],3a[3],...,na[n])
Alternatively, suppose
T'*(-U)=(1/t)T*D
T'* N =(1/t)T*S
Then dividing the two equations,
(T'*U)(T*S)+(T'*N)(T*D) =0
then
C = (S*N)D+(D*N)S
Jon G. wrote:
> Root Finder vii.
> by Jon Giffen
>
> The polynomial
>
> a[0]+a[1]t+a[2]t^2+a[3]t^3+ ... + a[n] = 0
>
> may be expressed as the dot product of the two vectors,
>
> T = (t,t^2,t^3,...,t^n) and
> N = (a[1],a[2],a[3],...,a[n]) then
>
> T*N+a[0]=0 is the nth degree polynomial
>
> Define,
>
> Q = (-a[0]/|N|^2)(a[1],a[2].a[3],..a[n])
>
> Suppose the vectors D,S,C are defined,
>
> D=( 1 , 2 , 3 ,...,n )
> S=(a[1],2a[2],3a[3],...,na[n])
> C=(S*N)D+(D*N)S
>
> then it can be shown that,
>
> |Q|
> T = Q +/- --- C
> |C|
>
>
> EXAMPLE:
>
> t^3+2t^2+t-4=0 t=1
> a[0]= -4
> a[1]= 1 a[2]=2 a[3]=1 N=(1,2,1) |N|^2=6 Q=(2/3)(1,2,1)
> |Q|=(2/3)6^(1/2)
> D=(1,2,3) S=(1,4,3) S*N=(1,4,3)*(1,2,1)=1+8+3=12
> D*N=(1,2,3)*(1,2,1)=1+4+3=8
> C=12(1,2,3)+8(1,4,3)=(12+8,24+32,36+24)=(20,56,60)=4(5,14,15)
> |C|=(4)446^(1/2)
>
> (2/3)6^(1/2)
> (t,t^2,t^3)=(2/3)(1,2,1)+/- ------------ (5,14,15)
> 446^(1/2)
>
>
> (t,t^2,t^3)=(2/3)(1,2,1)+/- 0.07732(5,14,15)
> (t,t^2,t^3)=(0.6666,1.3333,0.6666)+/-(0.3866,1.08254,1.1599)
>
> Selecting the first component,
>
> t = 0.6666+0.3866 = 1.0532 ~ 1
>
> then
>
> t^2 + 3t + 4
> ------------------------------
> t-1/ t^3 + 2t^2 + t - 4
> t^3 - t^2
> --------------------
> 3t^2 + t - 4
> 3t^2 -3t
> -------------
> 4t - 4
>
> and the remaining roots are
>
> -3+/-{9-16|^(1/2)
> t = ----------------- = -1.5+/-1.3229i
> 2
>
> Dividing t^3 by t^2,
>
> t^3 = 0.6666+1.1599 = 2.2656
> -----------------------------
> t^2 = 1.3333+1.0825 = 2.4158
>
> or
>
> t = 0.9378 ~ 1
>
>
> Jon Giffen
- Next message: Han de Bruijn: "Re: Cantor's diagonal proof wrong?"
- Previous message: Chris Menzel: "Re: Cantor's diagonal proof wrong?"
- In reply to: Jon G.: "Root Finder vii."
- Next in thread: C. Bond: "Re: Root Finder vii."
- Reply: C. Bond: "Re: Root Finder vii."
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|