Estimating Derivatives of Noisy Simulations
by J. Moré and S. Wild.
To appear in ACM Transactions on Mathematical Software, Vol. 38 (3), 2011.
[PDF from earlier preprint]

Do You Trust Derivatives or Differences?
by J. Moré and S. Wild.
Mathematics and Computer Science Division Preprint ANL/MCS-P2067-0312, 2012.
[PDF forthcoming]

Application problems,
comments, questions, and suggestions should be addressed to Stefan Wild at
.

Estimating Computational Noise

The following information (used in [1]) is provided to encourage
the estimation of computational noise in applications and to determine
strengths and limitations of the code.
We provide the following Matlab scripts for producing basic noise level estimates from data:

ECNdriver
[MATLAB], [fortran]
Estimates the noise given an array of function values from equally spaced points along a line

ECNoise
[MATLAB], [fortran]
Estimates the noise given an array of function values from equally spaced points along a line

ECNoise_space
[MATLAB], [fortran]
Estimates the noise given an array of function values from unequally spaced points along a line

Sample Problems

The following source files can be used to define the simulation-based problems in [1]:

mcfinance:function [MATLAB]
Code for MC evaluation of a finance calculation [Caflisch].

ptrace_L:function [MATLAB]
Code for evaluating a partial trace function.