cbrad

The findpeaks function in the MATLAB (verion 7.6.0.324) Signal Processing Toolbox finds local maxima or peaks in a vector.

>> v = [0 3 0 0 3 2];

>> [pks, locs] = findpeaks(v)
pks =
 3 3

locs =
 2 5

locs contains the locations or indices of the peaks in v (MATLAB indexing starts at 1).

For each column of a 256 x 5188 matrix we were calculating a simple moving average (smooth function) and then finding the peaks and it was taking a long time, approximately 19.6 seconds. We whipped up our own findpeaks function based on the following two lines:

>> s=v(2:end)-v(1:end-1);
>> loc=find((s(1:end-1)>0) & (s(2:end)<0)) + 1
loc =
 2 5

This calculates the pairwise differences between elements, i.e. the slopes and then looks for an up slope immediately followed by a down slope. Pretty simple and now the 5188 matrix columns are processed in approximately 1.3 seconds.

Been learning to use R for a stats subject. These links were useful.

• Quick-R


• R programming for those coming from other languages

Other things I have stumbled across along the way.

• R - Resources


• Rattle GUI for data mining in R


• SNA in R Talk, Updated with [Better] Video


• Revolutions blog

Update 1/11/2009
For completeness I should also have included An Introduction to R that comes bundled with R, as well as the official R Wiki. Also discovered RSeek.org as well as this MATLAB / R Reference.