Complicated for loops structuring

Hello, this is a question for the pros, and feel free to only answer if it interest you.

I have this code which helps me model data for each independent event.

% Event 1
sm1 = smcellere{1,:}; t1 = 1:length(sm1);SMF_1 = (1:0.3:SMO(1));good1 = sm1(~isnan(sm1)).';
figure(1)
plot(t1,sm1,'--ob');
hold on
for i = 1:length(SMF_1)
    for ii = 1:length(tau)
%         This is the model
        SM1 = (SMO(1)-SMF_1(i))*exp(-t1./tau(ii))+SMF_1(i);
        rmse1(i,ii) = sqrt(sum((SM1-good1).^2)/length(good1));
    end
end
[iii,jjj] = find(rmse1 == min(rmse1(:)));
smf1=SMF_1(iii);tau1 = tau(jjj);
sm1calc = (SMO(1)-smf1)*exp(-t1./tau1)+smf1;
plot(t1,sm1calc,'-or');
hold off

% Event 2
sm2 = smcellere{2,:}; t2 = 1:length(sm2);SMF_2 = (1:0.3:SMO(2));good2 = sm2(~isnan(sm2)).';
figure(2)
plot(t2,sm2,'--ob');
hold on
for ai = 1:length(SMF_2)
    for aii = 1:length(tau)
        SM1 = (SMO(2)-SMF_2(ai))*exp(-t2./tau(aii))+SMF_2(ai);
        rmse2(ai,aii) = sqrt(sum((SM1-good2).^2)/length(good2));
    end
end
[aai,aaj] = find(rmse2 == min(rmse2(:)));
smf2=SMF_2(aai);tau2 = tau(aaj);
sm2calc = (SMO(2)-smf2)*exp(-t2./tau2)+smf2;
plot(t2,sm2calc,'-or');
hold off

Now, the code is long, and this is only for 2 events. I have 34 events to go through, which I would be doing manually nonetheless if I can't figure out an algorithm to plot every event automatically. Furthermore, a working algorithm can help me cross check my answers.

This is what I'm trying to achieve by meshing everything into a single algorithm:

% Each event is represented as data in each row of cell
% Iterate through each event
for i = 1:length(smcellere)
    sm = smcellere{i,:};
%     The duration of event is denoted as t
    t = 1:length(sm);
%     SMO is the 1st data point for each row/event, and SMF is the interval
    SMF = (1:0.3:SMO(i));
%     Iterate through each data of SMF
    for ai = 1:length(SMF)
%         Iterate through each day in a year (tau)
        for aii = 1:length(tau) 
%             Plot the initial graph of observed data for each event
            figure, plot(t,sm,'--ob');
            hold on
%                   This is the exponential model. Run this model for each
%                   day in tau and each point in SMF
            SM = (SMO(i)-SMF(ai))*exp(-t./tau(aii))+SMF(ai);
            for gi = 1:length(SM)
                    for ki = t
%                     Calculate Root Mean Square Error to plot as line of
%                     best fit
                        rmse(ai,aii) = sqrt(sum((SM(gi)-sm(ki)).^2)/length(sm(ki)));
%                         THis is the best RMSE
                        [aai,aaj] = find(rmse == min(rmse(:)));
                        smf=SMF(aai);tau2 = tau(aaj);
%                         Calculate the model results
                        smcalc = (SMO(i)-smf)*exp(-t./tau2)+smf;
                    end
                end
        end
    end
%     Plot line of best fit
    plot(t,smcalc,'-or');
    hold off
end

The problem is, I can't do trial and error with this because my computer will just freeze due to bad coding.

Thanks in advance for any who'd like a challenge.

[Jonathan Cheong]

Cool, I've tried a few ways and am happy I'm finally one step closer. The code works perfectly, but only for the first 12 events. At event 13 it gives me an error:

"Index exceeds the number of array elements"

And I can't figure out why. I suspect the problem is marked with (*) , but do not know how to solve it.

smtmp = smcellere(:);
for i = 1:length(smtmp)
    sm = smtmp{i,:}.';
    t = 1:length(sm);
    SMF = (1:0.1:SMO(i));
    good = sm(~isnan(sm)).';
    figure, plot(t,sm,'--ob');
    hold on
    for ai = 1:length(SMF)
        for aii = 1:length(tau) 
            SM = (SMO(i)-SMF(ai))*exp(-t./tau(aii))+SMF(ai);
%             **rmse(ai,aii) = sqrt(sum((SM-sm).^2)/length(sm));**
            
        end
        
    end
    [aai,aaj] = find(rmse == min(rmse(:)));
%     if aai > length(SMF)
%         smf=SMF(length(SMF));tau2 = tau(aaj);
%     elseif aai < length(SMF)
    smf=SMF(aai);tau2 = tau(aaj);
%     end
    smcalc = (SMO(i)-smf)*exp(-t./tau2)+smf;
    plot(t,smcalc,'-or');
    hold off
end

Any input is much appreciated.

[darova]

I don't see rmse preallocation. Did you define size of the variable?

Answer 1

[Jonathan Cheong]

Thanks this works perfectly.

smtmp = smcellere(:);
for i = 1:length(smtmp)
    sm = smtmp{i,:}.';
    t = 1:length(sm);
    SMF = (1:0.1:SMO(i));
    figure, plot(t,sm,'--ob');
    hold on
    rmse = zeros(size(SMF,1),length(tau));
    for ai = 1:length(SMF)
        for aii = 1:length(tau) 
            SM = (SMO(i)-SMF(ai))*exp(-t./tau(aii))+SMF(ai);
            rmse(ai,aii) = sqrt(sum((SM-sm).^2)/length(sm));
        end
    end
    [aai,aaj] = find(rmse == min(rmse(:)));
    smf=SMF(aai);tau2 = tau(aaj);
    smcalc = (SMO(i)-smf)*exp(-t./tau2)+smf;
    plot(t,smcalc,'-or');
    hold off
end