home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!spool.mu.edu!yale.edu!ira.uka.de!math.fu-berlin.de!unidui!rrz.uni-koeln.de!Germany.EU.net!news.uni-bielefeld.de!math25!sillke
- From: sillke@math25.mathematik.uni-bielefeld.de (Torsten Sillke)
- Subject: discrete optimization (fixpoints of polynomials)
- Sender: news@hermes.hrz.uni-bielefeld.de (News Administrator)
- Message-ID: <BzDDqo.568@hermes.hrz.uni-bielefeld.de>
- Date: Wed, 16 Dec 1992 20:47:12 GMT
- Nntp-Posting-Host: math25.mathematik.uni-bielefeld.de
- Organization: Math MadHouse Bielefeld, Germany
- Keywords: discrete optimization, polynomial fixpoints
- Lines: 46
-
- Discrete Optimization:
-
- Find Polynomials
-
- n
- ---
- \ i n-i
- P (x) = > a x (1-x)
- n / i
- ---
- i=0
-
- with integer coefficients 0 <= a <= Binomial[n,i]
- i
-
- such that P_n(x) is not the identity and has as many as
- possible fixpoints in closed interval [0,1].
-
-
- For n = 1..5 you can get n different fixpoints in the interval [0,1].
- p_1 : [1,0] = [a0, a1]
- p_2 : [0,0,1] = [a0, a1, a2]
- p_3 : [0,3,0,1]
- p_4 : [0,0,6,2,1]
- p_5 : [0,2,0,10,3,1]
-
- The case n=6 is the first time, where you can get only n-1 different
- fixpoints in [0,1].
-
- For n=20 I managed to get 17 fixpoints. Can you beat this?
- What are optimalization procedures in this case?
- (Simulated Annealing with the neightbourhood 'change one coefficient
- a little bit' is no good idea, as almost all fixpoints will be
- complex typically.)
-
- You can construct polynomials with 0.2*n fixpoints. Can you find
- a better lower bound?
-
- This was a exercise in a 2nd year math course. We recieved one
- solution. Three students constructed (4 pages) a polynomial with
- fix points at k/(n-1). Then they showed that the coefficients
- are integers and in the allowed range. So they constructed
- at least n different fixpoints. But they missed that they had
- found the identity. They started their induction for n=3.
-
- Torsten Sillke
-