7.4 Interval Halving Method

Author: Miroslav Obradovic,  Index: 132/98
( mn98132@alas.matf.bg.ac.yu )

This is interval halving method for solving non-linear equations. Beginning with an interval [a0,b0]=[a,b] where the solution is localized, this method creates series of intervals [an,bn] (n=0,1,2,...) each being the half of the previous one, by moving only one interval boundary at a time, so to keep the solution embraced, i.e. x* ∈ [an,bn]. Criterion for solution localizing is that the function changes it's sign at interval boundaries, i.e. f(an)*f(bn) < 0.

The size of each interval can be calculated as: bn-an=(b-a)/2n, and it converges to zero when n->+oo. Solution approximation is the midpoint of interval, i.e. x(n)=(an+bn)/2 with error: |x*-x(n)|  <  (an+bn)/2  =  (b-a)/2n+1  <  epsilon.

</COMMENT> You must have Java 1.3 compatible browser to execute.