Lyuka and the Ostrowski method's for Root Seeking


A few threads back Hugh Steers pointed out a method for finding the roots of a nonlinear function by Russian mathematician Ostrowski. I checked Luyka's web page which discussed that method and presented the code for the algorithm. It did real well.

I further searched the Internet for articles about root-seeking method based on Ostrowski's method. I found several articles that further accelerated the convergence for that algorithm. A few articles were published by Chinese mathematician Zou (and colleagues) and offer somewhat elaborate enhancements to Ostrowski's method. These enhancements can solve for a root in TWO iterations (requiring 9 function calls) as compared to 6 iterations (and 12 function call) with Newton's method!. The fast Halley's method required 3 iterations and 9 function calls. The basic Ostrowski method took 3 iterations and 7 function call.

So thank you Lyuka for pointing me to Ostrowski's method. I really enjoyed learning about it and about its variants.

More gems in the root-seeking toolbox. Just when I thoght I learned about all the algorithms available, I get to learn about new and even more efficient ones!!




The Ostrowski's method is a kind of root finding algorithm that uses reverse-interpolation to approximate the root.

So, any other interpolation, such as Lagrange interpolation, can be used as a root finding algorithm.

If you use that of n=3, a recurrence equation

can be used instead of Ostrowski's

This can be tested replacing a line in _ost.c

t = (h * a - b) / (h - 1.0);

t  = a * e * f * (f - e) - b * d * f * (f - d) + c * d * e * (e - d);
t /= (f - e) * (f - d) * (e - d);

Though the convergence of the method shown above is almost quadratic (order of about 1.8) for a zero of multiplicity 1 in a neighborhood of the zero, it's NOT recommended as it tends to diverge when the guess is not near the zero.

IMHO, the most important thing as a root finding algorithm is not
the order of convergence, but the stability of convergence, i.e. ability to find a root with very few chance of divergence.



Possibly Related Threads...
Thread Author Replies Views Last Post
  A fast Bernoulli Number method for the HP Prime Namir 16 1,734 11-22-2013, 04:46 PM
Last Post: Namir
  [HP Prime] Using n-root symbol and exponent problem uklo 7 977 11-11-2013, 01:39 AM
Last Post: Alberto Candel
  A brand new calculator benchmark: "middle square method seed test" Pier Aiello 25 2,351 09-13-2013, 01:58 PM
Last Post: Pier Aiello
  Cubic root (-8) = 2 ? Gilles Carpentier 37 3,195 08-12-2013, 10:26 PM
Last Post: jep2276
  Downhill Simplex Method (Nelder Mead) for WP-34S Marcel Samek 0 299 07-07-2013, 03:05 AM
Last Post: Marcel Samek
  Square Root Simplifier for HP39gII Mic 4 690 03-11-2013, 08:25 AM
Last Post: Eddie W. Shore
  Seeking DIY4X update David Griffith 10 1,046 11-15-2012, 12:41 AM
Last Post: Eric Smith
  Cube root on standard calculator Thomas Klemm 22 2,212 11-09-2012, 06:11 AM
Last Post: Pierre
  Seeking new resources for TI NSpire Programming Namir 6 783 07-29-2012, 03:57 AM
Last Post: fhub
  New Quadratic Lagrangian Interpolation Method Namir 2 467 07-20-2012, 04:32 PM
Last Post: Namir

Forum Jump: