You are on page 1of 2

NUMERICAL METHODS

AKIRA KOBASHI

1. LINEAR ITERATIVE METHOD This simple method is designed for equations of the type x = g (x) Starting with a suitable approximation x0 to the root , we compute xi = g (xi ) or i = 0, 1, 2, ... and obtain the sequence x1 , x2 , x3 , .... If this sequence converges to , then from xi+1 = g (xi ), we have
i

lim xi+1 = lim g (xi )


i

or = g () . this implies is a root of x = g (x) 1.1. Condition for Convergence. If g(x) and its rst derivative g(x) are continuous and |g (x)| M < 1 in the interval of iteration then the sequence of titerates will converge for any starting value x0 in the interval. (The interval of iteration is the intervl in which all the iterattes, including the starting value x0 lie.) CONDITION. The condition on the derivative may be replaced by the less stringent 1 uniform Lipschits condition: |g (xi+1 ) g (xi )| M |xi+1 xi |, M < 1 PROOF. From xi = g (xi1 ) , we have |ei | = | xi | = g () g (xi1 ) M |xi1 | = M |ei1 | Thus, |ei | M |ei1 | M 2 |ei2 | M i |e0 | Hence we have convergence with M < 1. The order of the method is 1 and after the ith iteration the initial erroris reduced by a factor of M i . Obviously, the smaller the M is, the faster is the rate f convergence. The condition is sucient but not necesssary. That is to sy that the condition guarantees that the iterates will converge to the solution, but htere may be equations whre gx) do not satisfy the condition but nevertheless the iteration produces a valid solution. However, working with such g(x) one cannot be sure of obtaining a solution.
Date : April 1, 2014.
1stringent Laws, rules, or conditions that are stringent ar very strictly controlled or enforced

because they must be obeyed. EG There were no stringent rules applied to the huntingof

deer in the forest. EG He gave his sister one hundred dollars ermonth underthe stringent condition that she wuld never enter his house again. EG It may be that we need to introduce stringent controls.
1

AKIRA KOBASHI

1.2. Adaptatioon of the Result for Computation. he mathematicl resulat means that if M and |e0| were known and the evaluations of g (xi ) could be carried out exactly then could be approximated with the desired accuracy by stopping when M i |e0 | became less than the prescribed error . However, neither M nor |0 | will be known (because ] will not be known) and omputations will not be exact. So how should we know that a root has been found with the desired accuracy? 1.3. Error Estimate in a Computationally Useful Form. From xi1 = xi + xi xi1 we get, using xi = g () g (xi1 ) xi1 |xi |+|xi xi1 | |g ()g (xi1 )|+|xi xi1 | M |xi1 |+|xi xi1 | Hence, xi xi1 xi1 = |g () g (xi1 )| 1M and M xi |x xi1 | 1M since | xi | = |g () g (xi1 )| M | xi1 | 2. Example 5.5 PROBLEM: Compute a real root from f x) = x3 3x 5 = 0 using the iterative method. SOLUTION: Here we can rewrite x3 3x 5 = 0 as x = (x) = Asssume initial value of x is 2. x=2 for n in xrange(8): x=sqrt(3+5/x) print x.n() We can get the solution, 2.2790196026245. 3. SAGE Solutions Actually, in this case, we do not need to compute numerically. Command sage that saGe:nd root(xs-3x-5==0, 1, 3). Thats all. This solution is as easy as sin(x) = 10(x 1) Ask sage that sage: nd root(sin(x)==10*(x-1),-pi,pi). Thats all. We might not need to remember how to solve those dicult theories. 3+ 5 x

You might also like