Re: Solving For 0
- To: mathgroup at smc.vnet.net
- Subject: [mg25827] Re: Solving For 0
- From: "Paul Lutus" <nospam at nosite.com>
- Date: Sat, 28 Oct 2000 01:41:19 -0400 (EDT)
- References: <8t64tq$ei1@smc.vnet.net>
- Sender: owner-wri-mathgroup at wolfram.com
<bobmarley4u2c at my-deja.com> wrote in message news:8t64tq$ei1 at smc.vnet.net... > I noticed that when I use my Ti-89 or Mathematica 4.0, they solve for > the roots very fast. I tried to create program simliar to that in C++, > but I am having no luck. I tried the Iterative Method, which works, but > it is extremly long. Then I tried Newton-Raphson method, which I got > only to work for 3rd degree polynomial. It Newton-Rapson method just > gives me the same answers after the 3rd degree. In my long experience in programming, "divide and conquer" has been the best and most robust approach to root finding. Design a binary search that begins approaching the root and, on each pass through the root, reverses the sign and size of the remaining distance. When the result falls below a chosen epsilon, halt. You must also realize that no root finder will find all roots of a complex function on its own. You must have some idea of the locations of the roots, or be willing to let the algorithm search for a long time. > e-mail me at nganta82 at hotmail.com In general, it is better to read the newsgroup. This way, if someone offers defective advice, someone else will correct it. It's democracy at work. -- Paul Lutus www.arachnoid.com