[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference rusure::math

Title:Mathematics at DEC
Moderator:RUSURE::EDP
Created:Mon Feb 03 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2083
Total number of notes:14613

1823.0. "Convolution recurrence" by RTL::GILBERT () Wed Dec 08 1993 13:41

    What is a closed form solution to the recurrence:
    
    	T(0) = 1
    	        d
    	T(d) = Sum T(i-1)*T(d-i), d > 0
    	       i=1
T.RTitleUserPersonal
Name
DateLines
1823.1HANNAH::OSMANsee HANNAH::IGLOO$:[OSMAN]ERIC.VT240Fri Dec 10 1993 09:3833

How do we define "closed form" ?

I was thinking about this the other day.

Suppose I want the 6th Fibonacci number.

Well, I could do it in the "non-closed" form where I say

1+1 is 2, 1+2 is 3, 2+3 is 5, 3+5 is 8, 5+8 is 13


So the essence of "non-closed" seems to be that we do some sort of iterative
process to get our answer.

But then some mathematician comes along and presents the "closed form" for
Fibonacci numbers as

	       (1+sqr(5))^n - (1-sqr(5))^n
	f(n) = ---------------------------
		          sqr(5)* 2^n

However, it still seems like I have to do an iterative process to compute
Fibonacci numbers using this formula, since unless I use some sort of log tables
I'll end up computing the nth power by iteratively multiplying.

Why is the interative multiplying to compute nth power considered "closed form"
but the iterative adding of previous two terms considered "non-closed form" ?

/Eric


1823.2Closed Form is a relative term.CADSYS::COOPERTopher CooperFri Dec 10 1993 12:5215
    "Closed form" is not defined in any consistent fashion.  It depends
    on context.  A particular expression is considered of "closed form"
    only relative to some set of predetermined functions and operators.

    Generally it includes +, -, *, /, and "standard" transcendental
    functions such as log, exp, sin, hsin, etc. (Once you have log and exp
    you may as well have the power operator), plus some assorted others
    like sgn, and abs.  Depending on context, other functions may be
    included, such as the error function.

    What is always excluded is explicit use of indefinite numeric
    "quantifiers", such as sigma (sum), pi (product), and various
    integrals.

				    Topher
1823.33D::ROTHGeometry is the real life!Fri Dec 10 1993 20:2016
   I numerically experimented with this and found that

	 t(d)      2d - 1
        ------ = 4 ------
	t(d-1)     2d + 2

   So that t(d) should have the "closed form" continued
   product

	t(0) = 1

                    d    2*k - 1
	t(d) = 4^d PROD  -------, d > 0
                   k = 1 2*k + 2

   - Jim
1823.4a seedRTL::GILBERTMon Dec 13 1993 14:3817
    Thanks, Jim.  Jim also asked where this problem came up.
    T(d) is the number of ways to decompose a polynomial of degree d
    into an expression tree by recursively applying the form:
    	                   i
    	P (x) = P   (x) + x  R   (x),
    	 d       i-1          d-i
    
    where (as you'd expect)
    	                            d
    	P (x) = c  + c x + ... + c x ,
    	 d       0    1           d
    	                                i-1
    	Q   (x) = c  + c x + ... + c   x   , and
    	 i-1       0    1           i-1
    	                                d-i
    	R   (x) = c  + c   x + ... + c x   .
    	 d-i       i    i+1           d
1823.5RTL::GILBERTMon Dec 13 1993 15:107
    re .3:
    
    BTW, this simplifies to
    
                 (2d)!
    	T(d) = --------- = C(2d,d)/(d+1).
               (d+1)! d!