Services & Resources / Wolfram Forums
-----
 /
MathGroup Archive
2003
*January
*February
*March
*April
*May
*June
*July
*August
*September
*October
*November
*December
*Archive Index
*Ask about this page
*Print this page
*Give us feedback
*Sign up for the Wolfram Insider

MathGroup Archive 2003

[Date Index] [Thread Index] [Author Index]

Search the Archive

Re: Alternative to bivariate Taylor series

  • To: mathgroup at smc.vnet.net
  • Subject: [mg44679] Re: Alternative to bivariate Taylor series
  • From: Paul Abbott <paul at physics.uwa.edu.au>
  • Date: Fri, 21 Nov 2003 05:13:04 -0500 (EST)
  • Organization: The University of Western Australia
  • References: <bpd2a7$ce5$1@smc.vnet.net>
  • Sender: owner-wri-mathgroup at wolfram.com

In article <bpd2a7$ce5$1 at smc.vnet.net>, koshi man <man_koshi at yahoo.com> 
wrote:

> Is there any expansion to log(1+exp(x-y)) other than
> Taylor series in the integer powers of x and y ? 

There are many other possible expansions. One possibility is that 
Log[1+Exp[x]] is well approximated by truncating the following expansion:

  x UnitStep[x] - Sum[(-1)^i Exp[-i Abs[x]]/i, {i, 1, Infinity}]

Another is to use FindFit to find optimal values for the parameters a, 
b, c, d in the following ansatz:

 x UnitStep[x]+ a Exp[-b Abs[x]] + c Abs[x] Exp[-d x^2] 

> I'm trying to find expectation of above function given
> x, y are random variables with gaussian distributions
> of known  mean and variances.

For numerical parameters, you can use NIntegrate to compute these 
expectation values directly via a straightforward double integral.

Assuming x and y are independent, for the Normal distribution, you can 
reduce the computation of expectation values from a 2D integration to a 
1D integration (which I do not think reduces to a closed form for 
log(1+exp(x-y))). See the Notebook at

 http://physics.uwa.edu.au/pub/Mathematica/MathGroup/ExpectationValues.nb

> What about other bivariate expansions in Mathematica?

You can always try NIntegrate. Because your function depends linearly on 
a x + b y, for some distributions you will again be able to reduce the 
computation to a 1D numerical integration.

Also, have a look at "Mathematical Statistics with Mathematica" by Colin 
Rose and Murray D. Smith:

  http://store.wolfram.com/view/book/ISBN0387952349.str

Cheers,
Paul

-- 
Paul Abbott                                   Phone: +61 8 9380 2734
School of Physics, M013                         Fax: +61 8 9380 1014
The University of Western Australia      (CRICOS Provider No 00126G)         
35 Stirling Highway
Crawley WA 6009                      mailto:paul at physics.uwa.edu.au 
AUSTRALIA                            http://physics.uwa.edu.au/~paul


  • Prev by Date: Re: FindMinimum Problem
  • Next by Date: Re: [Integrate] Why two results of same eq. are different?
  • Previous by thread: Re: Alternative to bivariate Taylor series
  • Next by thread: web-based mathematical visualization and interactive geometry