Alternatives to 'datasample'

Hi!

I'm looking for an alternative to the datasample function. It's a part of the Statistics & Machine Learning toolbox. Any ideas? I want to be able to sample rows of data from tables and matrices and would like to have the choice of sampling with/without replacement.

Thanks for your ideas!

[Adam Danz]

Are you sampling with replacement or without replacement?

[Veena Chatti]

Hi Adam,

I am sampling with replacement. My dataset (of personally curated "ground truths") contains hundreds of samples, and I am using it to train a pattern recognition perceptron that needs 200,000 epochs at a reasonable learning rate.

Thanks,

Veena

[Veena Chatti]

I think I've found a way to do it by converting the dataset to an array, combining a loop with randperm, and choosing one row at a time, but ideally, if someone knows about a function that does it in a more straightforward way that I just haven't come across yet, that would be fabulous. I eventually want to share my perceptron with others who may not have the toolbox, and it would be nice not to require that they have to buy it just for this one function!

[Adam Danz]

randperm samples without replacement.

To sample with replacement, use randi.

[Veena Chatti]

Thanks. I think I have it now.

Answer 1

[Kiran Felix Robert]

Hi Veena,

Yes, randomperm (without replacement) and randi (with replacement) can be used as an alternative.

Furthermore, if the dataset has been converted to either a matrix or an array format, the rows can be sampled without using a loop as shown in the example as follows,

X = rand(50,10); % Assume X is the dataset 
SampledRows = X(randi(50,[200,1]),:); % 200 Sample rows (with replacement) 

Kiran Felix Robert

[Veena Chatti]

Thank you Kiran Felix Robert!