RMSE of sequences in cell arrays with different length

In the Sequence-to-Sequence Regression Using Deep Learning example the rmse is caclulated this way

for i = 1:numel(YTest)
    YTestLast(i) = YTest{i}(end);
    YPredLast(i) = YPred{i}(end);
end
figure
rmse = sqrt(mean((YPredLast - YTestLast).^2))

in this case we are interested only on the last element so YPredLast - YTestLast is one to one element substraction

IN the general case that i am interested on the whole output sequence, what will be the substraction??

YTest - Ypred cannot be performed as these are cell arrays.

how can we calculate the rmse from the whole sequencies?

[madhan ravi]

You use (i) and you say it’s a cell array ?? No sign of pre-allocation whatsoever.

[sotiraw sotiroglou]

Please read more carefully the questions before you write your answers..

The code provided on the example is for a specific example , where the Ytest is a cell array. And YtestLast is not a cell array. So Ytest is a cell array and i am asking how to caclulate rmse on Ytest, if we are interested on a simular example that we do not only want the last element of it, but the whole sequence. I hope this answers to you question???

As for the pre-allocation , i just used an official matlab example. And their code , do you have any comment on the question i ask? thanks

Answer 1

[Star Strider]

I am not certain what final result you want.

This should get you started:

subt = cellfun(@minus, YTest, YPred, 'Uni',0);                  % Vector Differences
RMSE = cellfun(@(x)sqrt(mean(x.^2)), subt, 'Uni',0);            % RMSE Of Vector Differences

You can use the cell2mat function to convert the ‘RMSE’ vector to a double vector, and then do any other necessary calculations on it.

Experiment to get the result you want.