Example 6.  Compare the Horner method and ordinary method of polynomial evaluation for
[Graphics:Images/HornerMod_gr_179.gif]  

Solution 6.

Construct a the polynomial and two version of Horner's method.

Aside.  This polynomial has known roots  [Graphics:../Images/HornerMod_gr_180.gif].  

[Graphics:../Images/HornerMod_gr_181.gif]


[Graphics:../Images/HornerMod_gr_182.gif]


[Graphics:../Images/HornerMod_gr_183.gif]



[Graphics:../Images/HornerMod_gr_184.gif]

For illustration we can evaluate [Graphics:../Images/HornerMod_gr_185.gif].

[Graphics:../Images/HornerMod_gr_186.gif]



[Graphics:../Images/HornerMod_gr_187.gif]

[Graphics:../Images/HornerMod_gr_188.gif]

 

 

We are done.

Aside.  Consider the computational time for vector arguments for evaluating  P(x)  and the nested multiplication form  Q(x)  and Horner form  H(x).  The latter two take roughly  1/2  the time to do the evaluations

[Graphics:../Images/HornerMod_gr_189.gif]


[Graphics:../Images/HornerMod_gr_190.gif]

 

 

We are really done.

Aside.  The product form of the polynomial,  [Graphics:../Images/HornerMod_gr_191.gif]  is the "best computational form."  We can compare the "error" with it and the other three forms of this polynomial.  

[Graphics:../Images/HornerMod_gr_192.gif]


[Graphics:../Images/HornerMod_gr_193.gif]

[Graphics:../Images/HornerMod_gr_194.gif]

[Graphics:../Images/HornerMod_gr_195.gif]

[Graphics:../Images/HornerMod_gr_196.gif]

[Graphics:../Images/HornerMod_gr_197.gif]

[Graphics:../Images/HornerMod_gr_198.gif]

An analysis of the error has been done by G. W. Stewart in the article:
Error Analysis of the Algorithm for Shifting the Zeros of a Polynomial by Synthetic Division  
G. W. Stewart  
Mathematics of Computation, Vol. 25, No. 113. (Jan., 1971), pp. 135-139, Jstor.  

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

(c) John H. Mathews 2004