Results 1 to 5 of 5

Math Help - Expansion of taylor's formula in 3 variables?

  1. #1
    Newbie
    Joined
    Dec 2007
    Posts
    3

    Expansion of taylor's formula in 3 variables?

    I am having difficulties figuring out how to write the full expansion of taylor's formula in 3 variables (x,y,t), for degrees 2, 3 and even higher degree terms.

    It is for a master thesis project in computer vision.

    The remainder term(s) is not necessary, but would be nice to have as well.

    This is taylor's formula expansion for f(x) containing up to first degree terms only (with remainder R_1):

    f(b)=f(a)+f_x(a)(b-a) + R_1

    And, if we exchange some variables: b=x+\Delta x and a=x, we have:

    f(x + \Delta x)=f(x)+f_x(x) \Delta x + R_1

    Using this notation instead, this is my taylor's formula expansion for f(x,y,t) up to first degree terms (partial derivatives in x, y, and t):

    f(x + \Delta x, y + \Delta y, t + \Delta t)=
    f(x,y,t) + f_x(x,y,t) \Delta x + f_y(x,y,t) \Delta y + f_t(x,y,t) \Delta t + \text{remainder terms...}

    Now I need the expansion in all three variables, for higher degrees (second, third, perhaps four) How to do this? I have found functions for this, but these are using multi-index vectors for indexing, and unfortunately I have been unable to wrap my head around such notation.

    As a side note, I currently expect mixed/crossed partials to be equal for function f(x,y,t) (which I do not know analytically).

    Thanks,
    Intel4004
    Follow Math Help Forum on Facebook and Google+

  2. #2
    Global Moderator

    Joined
    Nov 2005
    From
    New York City
    Posts
    10,616
    Thanks
    10
    Let S be a non-trivial open convex subset of \mathbb{R}^n. Then suppose f: S \mapsto \mathbb{R} is a \mathcal{C}^k function.
    Then,
    f(\bold{a}+\bold{h}) = \sum_{|\alpha|\leq k} \frac{\partial ^{\alpha} f(\bold{a})}{\alpha !}\bold{h}^{\alpha} + R_k(\bold{h}).
    Here we are using multi-index notation
    Follow Math Help Forum on Facebook and Google+

  3. #3
    Newbie
    Joined
    Dec 2007
    Posts
    3
    Quote Originally Posted by ThePerfectHacker View Post
    Let S be a non-trivial open convex subset of \mathbb{R}^n. Then suppose f: S \mapsto \mathbb{R} is a \mathcal{C}^k function.
    Then,
    f(\bold{a}+\bold{h}) = \sum_{|\alpha|\leq k} \frac{\partial ^{\alpha} f(\bold{a})}{\alpha !}\bold{h}^{\alpha} + R_k(\bold{h}).
    Here we are using multi-index notation
    Exactly! That is the formula that I have tried to understand. Actually, if you could tell me how the summation is supposed to iterate with \alpha, I believe I could go a long way. That is the only thing that stops me.

    As an iterator value in the summation, one uses the length/size |\alpha|=(\alpha_1 + \cdots + \alpha_n) ...right? But what is the limit of the sum? What terms doees the sum produce?

    Thanks.

    And, btw, in any event it applies to where you are, merry xmas

    Intel4004
    Last edited by intel4004; December 24th 2007 at 12:25 AM.
    Follow Math Help Forum on Facebook and Google+

  4. #4
    Newbie
    Joined
    Dec 2007
    Posts
    3
    The formula is familiar to the last one in this PDF:

    http://www.math.washington.edu/~solomyak/TEACH/425/05/taylor.pdf

    So as I understand, \alpha_n in \alpha, is an integer, describing how many times we differentiate with respect to variable number n?

    And since f is a \mathcal{C}^k function, it is k differentiable, right? Does that mean that all crossed partials (up to k'th order) can be regarded equal?

    And, btw, what do you mean?

    Like this?:

    f(\bold{a}+\bold{h}) = \sum_{|\alpha|\leq k} \left ( \frac{\partial ^{\alpha} f(\bold{a})}{\alpha !}\bold{h}^{\alpha} \right ) + R_k(\bold{h})<br />

    or:

    f(\bold{a}+\bold{h}) = \sum_{|\alpha|\leq k} \left ( \frac{\partial ^{\alpha} f(\bold{a})}{\alpha !}\bold{h}^{\alpha} + R_k(\bold{h}) \right ) <br />
    Follow Math Help Forum on Facebook and Google+

  5. #5
    Global Moderator

    Joined
    Nov 2005
    From
    New York City
    Posts
    10,616
    Thanks
    10
    The one I wrote is the correct one.


    Suppose f: \mathbb{R}^2\mapsto \mathbb{R} is \mathcal{C}^3 on a open convex set.
    Then,
    f+\frac{f_x}{1!}+\frac{f_y}{1!}+\frac{f_{xx}}{2!}+  \frac{2f_{xy}}{2!}+\frac{f_{yy}}{2!}+\frac{f_{xxx}  }{1!}+\frac{3f_{xyy}}{3!}+\frac{3f_{xxy}}{3!}+\fra  c{f_{yyy}}{3!}

    The reason why we divide f_{xxy} by 3! is because of the three partial derivatives but there is a 3 in the numerator is because we can permute f_{xxy} as f_{xyx}, f_{yxx} and all of them are the same.

    That was the Taylor's theorem for 2 variables. The same situation for 3 or more but it gets messier. The general pattern is the multi-index formula.
    Follow Math Help Forum on Facebook and Google+

Similar Math Help Forum Discussions

  1. Taylor formula for function of 2 variables
    Posted in the Calculus Forum
    Replies: 2
    Last Post: November 8th 2009, 12:08 PM
  2. Taylor Expansion
    Posted in the Calculus Forum
    Replies: 2
    Last Post: July 28th 2009, 12:56 AM
  3. Taylor expansion
    Posted in the Differential Geometry Forum
    Replies: 1
    Last Post: June 10th 2009, 08:05 AM
  4. Taylor expansion
    Posted in the Calculus Forum
    Replies: 1
    Last Post: May 5th 2008, 10:55 AM
  5. Taylor Expansion....
    Posted in the Calculus Forum
    Replies: 4
    Last Post: November 13th 2007, 09:23 PM

Search Tags


/mathhelpforum @mathhelpforum