# Modified Newton's Method

I don't understand how to find the rate of convergence of a root-finding method.

My professor drew out the standard Newton's method and explained that it has quadratic convergence. Then he asked how to find the rate of convergence if the method was modified to $$x_{n+1}=x_n-\frac{f(x_n)}{f'(x_0)}$$

Is it still converging quadratically? How do I best show this mathematically?

UPDATE

I found a place that showed the error definition of the original method. I tried to substitute my change into that proof, but it relied on Taylor Expansions for the prime term, which wasn't present in the same form for my problem... Dead End.

Then I made a drawing using a basic $f(x) = x^2$ graph and realized that since the slope of the tangent line to the x-axis is constant, it produces similar triangles.

Then looking at the ratios of the areas of the triangles from each iteration, I should be able to derive a rate of convergence knowing that the areas of similar triangles with scale factor $a:b$ are proportionate to $a^2:b^2$.

Here is one approach to understanding Newton's method that generalizes easily to your situation. Let $n(x)=x-f(x)/f'(x)$. Newton's sequence approximating a root is defined recursively by $x_{k+1}=x_k-f(x_k)/f'(x_k)$ or $x_{k+1}=n(x_k)$. Thus, if $c$ is a root of $f$, we are interested in the difference $n(x)-c$, when $x$ is close to $c$. This can be estimated using a series expansion of $n$ about $c$:

$$n(x) \approx c+\frac{f''\left(c\right)}{2 f'\left(c\right)}\left(x-c\right){}^2+O\left(\left(x-c\right){}^3\right).$$

From here, it's pretty easy to see that the difference between $n(x)$ and $c$ is proportional to $(x-c)^2$, i.e. you expect quadratic convergence of the sequence defined recursively by $x_k=n(x_{k-1})$.

Now, try the same thing with $n(x)=x-f(x)/f'(x_0)$, where $x_0$ is the fixed first term in your sequence. Again, expanding $n$ about $c$, where $f(c)=0$, we get

$$n(x) \approx c+\left(1-\frac{f'(c)}{f'\left(x_0\right)}\right)(x-c) - \frac{f''(c)}{2 f'\left(x_0\right)}(x-c)^2 +O\left((x-c)^3\right).$$

Quadratic convergence is now lost due to the first order term.

We can use this series to help formulate some concrete examples. The critical issue is the absolute value of $1-f'(c)/f'(x_0)$:

• If $|1-f'(c)/f'(x_0)|>1$, we have divergence,
• If $0<|1-f'(c)/f'(x_0)|<1$, we have linear convergence,
• If $1-f'(c)/f'(x_0) = 0$, we have quadratic convergence.

Of course, those statements all assume that $x_0$ is sufficiently close to $c$.

Now, consider examples of the form $f(x)=x^2-c^2$, which has a root at $x=c$. Then, our series expansion becomes

$$n(x) \approx c + \left(1-\frac{c}{x_0}\right)(x-c) + \frac{1}{2x_0}(x-c)^2+ O\left((x-c)^3\right).$$

Comment: In fact, the second order approximation is exact for this family of functions, but that's not necessary for this approximation technique to work.

It's now very easy to produce specific types of behavior in this family. Whenever, $0<c<x_0$, for example, we have $0<1-c/x_0<1$ so we are guaranteed linear convergence. Even more specifically, if $c=2$ and $x_0=4$, then $1-c/x_0 = 1/2$ and this modified method generates a sequence whose difference from the root $c=2$ is cut about in half with each iterate. On the other hand, if $c=4$ and $x_0=1$, then $1-c/x_0=-3$ and we'll generate a divergent sequence.

Finally, consider $f(x)=2 x - x^3 + x^5$. Note that $c=0$ is a root of $f$ and that $f'(c)=2$. Furthermore, $f'(\sqrt{3/5})=2$. Thus, if we start this modified Newton's method at $x_0=\sqrt{3/5}$, then we might expect quadratic convergence. In fact, we get even better as $n(x) = (x^3-x^5)/2$ and any $x_0$ in $[0,1]$ leads to a sequence with cubic convergence.

Geometrically speaking, Newton's method finds the tangent of a function at a point, draws it down to the $x$-axis to locate a new point, and repeats.

Now, you can easily see how this works, because $f'(x_n)$ is the slope of the tangent line of $f$ at $x_n$.

But what happens if, instead of drawing the line down to the $x$-axis using the tangent of the function at that point, you draw a line with a fixed slope (namely, with slope $f'(x_0)$? It should be easy to see that the method will probably diverge.

• This doesn't guarantee divergence, however. Consider something like $f(x) = e^{-x}-1$, and take $x_0 = -20$. This gives you a really steep slope. You'll probably get to the root using this approach, but it's going to take you a long time because you never project forward very far at any given step. – Emily Jan 20 '14 at 16:51
• yes, this is my current understanding of the problem. I need to show what the actual rate of convergence is in some informal proof. I've been looking on the wiki at the quadratic convergence of the original method, but it's hard to see how the change effects the proof. – Neurax Jan 20 '14 at 17:09
• Think about the conventional Newton's method, where you draw tangent lines to the horizontal axis. This convergence is quadratic (normally). Now, think about the triangles formed by these tangent lines (that is, draw the tangent from $f(x_n)$ to the x-axis, then draw a line from $(x_n,0)$ straight up to $(x_n,f(x_n))$). Their area decreases quadratically. Now, what happens to the area of the triangles with a fixed slope forming the hypotenuse? – Emily Jan 20 '14 at 17:13

If you are after a proof, I would recommend tracing the steps of the proof here to see what happens with the modified version.

In simple situations, like $f(x) = x^2 - 2$, the modified version still converges.

• I've tried to use that proof to find a new convergence. It doesn't lead me to anything I recognize. I found a website that said to look at error $\epsilon_{n+1} = f(\epsilon_n)$ and find something recognizable in that function. Keeping the slope constant means that you are shifting your solution by a scalar amount of the functional value at the previous value. I don't know what that means for the convergence in formal math terms. – Neurax Jan 20 '14 at 19:53