@Bharat: It can be proven that Newton's method for square root, x(n+1) = 
(x(n) + a / x(n))/2, always converges to sqrt(a) if a > 0 and x(0) > 0. It 
is not difficult, so let e(n) = x(n) - sqrt(a), the error in the n-th 
iteration, and find the recurrence for e(n+1) in terms of e(n) and a. You 
should get e(n+1) = e(n)^2 / (2*(e(n) + sqrt(a))). You can then observe 
that if x(0) > 0, then e(1) >= 0, and if e(n) > 0 then 0 < e(n+1) < e(n) / 
2, proving convergence. You can further observe that if e(n) is "small," 
then convergence is quadratic: e(n+1) = O(e(n)^2).
 
Dave

On Monday, January 14, 2013 2:09:22 AM UTC-6, bharat wrote:

> @Don : But, newton's formulae doesn't always converge.. if our guess is 
> bad enough, it may diverge also.
>
> On Tue, Jan 8, 2013 at 8:30 PM, Don <dond...@gmail.com <javascript:>>wrote:
>
>> Sure,
>>
>> Let's try two examples:
>> x=1,038,381,081
>>
>> The last digit is 1, so continue
>> Now start with y=10,000 because that is half as many digits as x.
>> y0 = 10,000
>> y1 = 56919
>> y2 = 37581
>> y3 = 32605
>> y4 = 32226
>> y5 = 32226
>> y6 = 32223
>> y7 = 32223
>>
>> Y6=Y7 so you are done. Now square y7 giving 1,038,321,729. That is not
>> equal to x, so x is not a perfect square.
>>
>>
>> Second case
>> x=1,038,579,529
>> Last digit is 9, so continue.
>> y1 = 10000
>> y2 = 56928
>> y3 = 37585
>> y4 = 32608
>> y5 = 32229
>> y6 = 32227
>> y7 = 32227
>>
>> 32227^2 = x, so x is a perfect square.
>>
>> Don
>>
>>
>> On Jan 5, 8:08 am, bala bharath <bagop...@gmail.com> wrote:
>> >  @Don,
>> >                Can u explain with an Example...?
>> >
>> > With regards,
>> >
>> >  Balasubramanian Naagarajan,
>> >
>> >                                                              Chettinad
>> > College of Engg & Tech.
>> >
>> >
>> >
>> >
>> >
>> >
>> >
>> > On Sat, Jan 5, 2013 at 1:48 PM, Malathi <malu....@gmail.com> wrote:
>> > > Check this. It might help.
>> >
>> > >http://www.johndcook.com/blog/2008/11/17/fast-way-to-test-whether-a-n.
>> ..
>> >
>> > > On Sat, Jan 5, 2013 at 1:47 AM, Don <dondod...@gmail.com> wrote:
>> >
>> > >> start with a guess y. If you can arrange for y to be about half the
>> >
>> > > --
>> >
>> > > With Regards,
>> > >    Malathi
>> >
>> > > --
>>
>> --
>>
>>
>>
>

-- 


Reply via email to