[seqfan] Re: Looking for a simpler pattern for that matrix
Gottfried Helms
helms at uni-kassel.de
Fri May 9 05:08:54 CEST 2014
Seqfans, dear Alexander,
In answering Alexander's request for more detailed explanation
I'll restate the computation/decomposition of A in more detail here.
We'll denote the given (infinite) matrix of coefficients A
A = (1)
2 . . . . .
-5 3 . . . .
10 -22 4 . . .
-17 108 -57 5 . .
26 -432 504 -116 6 . ...
-37 1520 -3510 1600 -205 7 ...
...
and their elements as a{r,c} for row- and column-index, the indices
beginning at 1.
The Pascalmatrix P
P= (2a)
1 . . . . . . ...
1 1 . . . . . ...
1 2 1 . . . . ...
1 3 3 1 . . . ...
1 4 6 4 1 . . ...
1 5 10 10 5 1 . ...
1 6 15 20 15 6 1 ...
...
containing the binomial-coefficients p{r,c} = b(r-1 , c-1).
The n'th power of the Pascalmatrix, P^n, has the elements
p^(n){r,c} = n^(r-c) * b(r-1,c-1) (2b)
Then I gave the basic scheme, with some "magic" coefficients in
the column-vectors on the rhs
A_{,1} = P^1 * [2,3, 2, 0, ... ] (3)
A_{,2} = P^2 * [0,3,10,12, 0, ... ]
A_{,3} = P^3 * [0,0, 4,21, 36, 0, ,... ]
A_{,4} = P^4 * [0,0, 0, 5, 36, 80, 0, ,... ]
A_{,5} = P^5 * [0,0, 0, 0, 6, 55,150, 0, ,... ]
We rewrite the above columnvectors of coefficients as matrix M
M = (4a)
2 . . . . . . ...
3 3 . . . . . ...
2 10 4 . . . . ...
0 12 21 5 . . . ...
0 0 36 36 6 . . ...
0 0 0 80 55 7 . ...
0 0 0 0 150 78 8 ...
... ... ... ... ... ... ... ...
where
m{c ,c} = ( c+1) * c^0 (4b)
m{c+1,c} = (2c+1) * c^1
m{c+2,c} = ( c+1) * c^2
and all other entries are zero (4c)
--------------------------------------------------------------------------------
Then
c+2
abs( a{r,c} ) = Sum p^(c){r,k} * m{k,c} (5a)
k=c
c+2
abs( a{r,c} ) = Sum c^(r-k) * b(r-1,k-1) * m{k,c} (5b)
k=c
The sum notation and elements m expanded :
abs( a{r,c} ) = c^(r-c ) * b(r-1,c-1) * ( c+1)*1 (5c)
+ c^(r-c-1) * b(r-1,c ) * (2c+1)*c
+ c^(r-c-2) * b(r-1,c+1) * ( c+1)*c^2
and one more simplification gives
====================================================================
the complete evaluation for the signless a{r,c}:
abs( a{r,c} ) = c^(r-c) * ( (c+1)*( b(r-1,c-1) (6a)
+b(r-1,c )
+b(r-1,c+1))
+ c* b(r-1,c )
)
and finally introducing the signs
s = (-1)^(r-c) (6b)
a{r,c} = s * abs(a{r,c}) (7)
=====================================================================
=====================================================================
Example:
abs( a{4,3} ) = 3^1 * ( 4 *( b(3,2) + b(3,3)+ b(3,4) )+ 3*b(3,3))
= 3 * ( 4 *( 3 + 1 + 0 ) + 3*1 )
= 3*19
= 57
The sign is
(-1)^(4-3) = -1
so
a{4,3} = -57
which matches the entry in the matrix A.
Gottfried
Am 07.05.2014 02:28 schrieb Gottfried Helms:
> Dear Alexander / dear Frank -
>
> thank you for your help! Meanwhile I tried it myself hard and found
> a solution which can be nicely expressed in terms of row- and column-
> indexes.
>
More information about the SeqFan
mailing list