MATH 175: Numerical Analysis II Lecturer: Jomar Fajardo Rabajante IMSP, UPLB 2nd Sem AY 2012-2013 3rd Method: Regula Falsi/False Position/Inverse Linear Interpolation (also a bracketing method) We will still use the Intermediate Zero Theorem – The function should be continuous – The function values of the two endpoints of the line should be of opposite signs rd 3 Method: Regula Falsi y-axis ( x1 , f ( x1 )) x-axis ( x3 ,0) Approximate root ( x2 , f ( x2 )) rd 3 Method: Regula Falsi We will use inverse linear interpolation (in slope-intercept form): x2 x1 x x1 ( f ( x) f ( x1 )) f ( x2 ) f ( x1 ) To get x3, set f(x3)=0 (from the zero of the line): x2 x1 x3 x1 f ( x1 ) f ( x2 ) f ( x1 ) rd 3 Method: Regula Falsi y-axis ( x1 , f ( x1 )) Set new x1 = old x3 ( x3 ,0) New Approximate root x-axis ( x3 ,0) Old Approximate root ( x2 , f ( x2 )) rd 3 y-axis Method: Regula Falsi ( x1 , f ( x1 )) ( x3 ,0) Old Approximate root x-axis ( x3 ,0) New Approximate root ( x2 , f ( x2 )) Set new x2 = old x3 rd 3 Method: Regula Falsi y-axis 3rd 1st 2nd x-axis rd 3 Method: Regula Falsi Try this at home: create a flowchart of the Regula Falsi method Hint: use bisection algorithm, however change the formula for x3 to x2 x1 x1 f ( x2 ) x2 f ( x1 ) x3 x1 f ( x1 ) f ( x2 ) f ( x1 ) f ( x2 ) f ( x1 ) rd 3 Method: Regula Falsi • Like bisection method, Regula Falsi is guaranteed to converge to the root (assuming IZT is met). • The rate at which this method converges will depend on how nearly linear f(x) is near its zero. If f(x) is sufficiently differentiable then it is well approximated by a straight line over small intervals. • Regula Falsi is often (not always) faster than bisection method, but still the order of convergence is linear. rd 3 Method: Regula Falsi • We can use abs(x3,k-x3,k-1)<tol as our stopping criterion but not (b-a)/2k<tol (where [a,b] is the initial bracket) since the width of the bracket may not converge to zero like in bisection. • But a better stopping criterion is (Let x3,k be the approximate root at iteration k, and λ be the asymptotic error constant. Invoking the error evolution equation: errork errork 1 ) 1 where x3,k x3,k 1 tol x3,k x3,k 1 x3,k 1 x3,k 2 There’s an improved Regula Falsi method called the Modified Regula Falsi. y-axis 1/2 2nd 1st 3rd x-axis Bisection, Regula Falsi and Modified Regula Falsi methods are called Interval/Bracketing Methods. Next topics: Iterative methods, such as secant method, Newton’s method, fixed point iteration, Mϋller’s method, Bairstow’s method Usually, faster methods require more assumptions and offer fewer guarantees.