How to pre-allocate my loop for speed?

Hi,

I have a script to help process my collected altimeter data to derive seagrass heights. My script isnt quite working as one of my loops doesn't want to run properly and has an error of exceeding matrix bounds.

This is the loop that doesn't work: It is finding peaks above the seabed which it assumes are seagrass and the lower peaks as the seabed. I hope this makes sense.

for i = 1:length(ping_time)
    [pks,locs] = findpeaks(bin_corr(:,i),'NPeaks',3,'MinPeakDistance',3,'MinPeakProminence',4); % tuning units are distance = bin number, prom = correlation value
    if length(locs) == 1
        pk_seagrass(i) = NaN;
        pk_seabed(i) = bin_depth(locs);
    elseif length(locs) == 2
        pk_seagrass(i) = bin_depth(locs(1));
        pk_seabed(i) = bin_depth(locs(2));
    elseif length(locs) == 3
        pk_seagrass(i) = bin_depth(locs(1));
        pk_seabed(i) = bin_depth(locs(3));
    elseif length(locs) == 0
        pk_seagrass(i) = NaN;
        pk_seabed(i) = NaN;
    end
end 

Any help or suggestions are much appreciated! Thank you :)

[Jan]

Please post a copy of the complete error message. It contains important information and sharing thos makes it easier to help you.

Answer 1

[Star Strider]

I am not certain what the problem is, however the preallocation would be straightforward:

pk_seagrass = NaN(1,length(ping_time));
pk_seabed = NaN(1,length(ping_time));

This produces row vectors for both variables. Put them just above the for call.

.

[Daryna Butash]

Hi! Thank you very much for answering! I actually found the problem and it was nothing to do with the code and it actually worked very well, I realised that it was me using different editions of matlab (2020a and a 2021a one) with the latest one having no problems. Very glad to have it working, thank you for your time!

[Star Strider]

As always, my pleasure!

Note that preallocating is always appropriate. In my experience, it results in about a 20% improvement in execution speed. This is not always obvious with small arrays, however can reduce the processing time for large arrays by several minutes, even in computers with SSDs.

.