 
 
 
 
 
 
 
  
We use as example the Wilkinson polynomial of degree  where
 where  :
:
 
 the coefficient of
 the coefficient of  is 78. But if
we modify this coefficient by
 is 78. But if
we modify this coefficient by  there is a big change in the
roots, 4 of them becoming complex [4].
The general procedure leads to reasonable 
accurate result up to
 there is a big change in the
roots, 4 of them becoming complex [4].
The general procedure leads to reasonable 
accurate result up to  . At
. At  although Kantorovitch theorem has determined interval solutions that
indeed contain all the solutions,
Newton method is unable to provide an 
accurate estimate of this root due to numerical errors.
although Kantorovitch theorem has determined interval solutions that
indeed contain all the solutions,
Newton method is unable to provide an 
accurate estimate of this root due to numerical errors.
For  and if we are looking for the roots in the interval
[0,2] the computation time is 90ms, for
 and if we are looking for the roots in the interval
[0,2] the computation time is 90ms, for  190ms and 330ms for
 190ms and 330ms for  .
For the fast algorithm these times are: 10ms, 20ms, 30 ms
Note that the best classical solving algorithm start to give
inaccurate results for
.
For the fast algorithm these times are: 10ms, 20ms, 30 ms
Note that the best classical solving algorithm start to give
inaccurate results for  (between 12.5 and 18.5 the interval
analysis algorithm finds the roots 
13.424830, 13.538691, 15.477653, 15.498664, 17.554518, 17.553513)
 and give imaginary roots for
 (between 12.5 and 18.5 the interval
analysis algorithm finds the roots 
13.424830, 13.538691, 15.477653, 15.498664, 17.554518, 17.553513)
 and give imaginary roots for  .
.
 
 
 
 
 
 
