lsqcurvefit not working properly
Hi all, I have multiple sets of data that are to be fitted by my fitting function. Since these data sets are kind of related, what I did was to use an inital guess to find the parameter values for the first dataset, and then use the parameter values as the guess for the second dataset so on and forth. However, the fitting was really bad. May I ask if there is another way around this kind of problem involving fitting of multiple datasets?
tic
%% Preparation
clear; clc
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv"); % insert file path within parenthesis
%% Preamble
% Fundamental constants
h = 4.0135667696*10^-15; % units: eV/ Hz
c = 3*10^8; % SI units
kB = 8.617333268*10^-5; % units: eV/ K
% Data
wavelength = data(1:end, 1); % contains all of the wavelengths
delay_t = data(1, 1:end); % conatains all of the delay times
E = (h*c)./(wavelength*10^-9); % contains all of the probe energies
% Range of data needed
Range_E = E>=1.5 & E<=2.2;
Range_W = wavelength >= (h*c)/(2.2*10^-9) & wavelength <= (h*c)/(1.5*10^-9);
Range_T = delay_t>=0.5 & delay_t<=1000;
E_p = E(Range_E); % probe energies for scatterplot
w_p = wavelength(Range_W);
delay_T = delay_t(Range_T);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
Range_Wfit = w_p >= (h*c)/(2.2*10^-9) & w_p <= (h*c)/(1.62*10^-9);
E_fit = E_p(Range_Efit); % probe energies for fitting
W_fit = w_p(Range_Wfit);
t_min = min(delay_T);
t_max = max(delay_T);
w_min = min(w_p);
w_max = max(w_p);
[row1, col1] = find(data == w_min);
[row2, col2] = find(data == w_max);
[row3, col3] = find(data == t_min);
[row4, col4] = find(data == t_max);
% New cleaned up data
data_new = data(row1:row2, col3:col4);% new data containing required delta A
% for n = 1:length(delay_T)
% delta_Abs(:,n) = -1*data_new(:,n);
% delta_Abs_norm(:,n) = delta_Abs(:,n)/max(delta_Abs(:,n));
% delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
% delta_Abs_norm_fit(:,n) = delta_Abs_fit(:,n)/max(delta_Abs_fit(:, n));
% % plot command
% plot(E_p, delta_Abs_norm)
% xlabel(‘Probe Energy (eV)’)
% ylabel(‘Normalised Delta A (a.u.)’)
% legend(‘Experimental Data’)
% end
% Fitting function: Maxwell-Boltzmann distribution
function F = MB(y, e_fit)
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv");
h = 4.0135667696*10^-15;
c = 3*10^8;
kB = 8.617333268*10^-5;
delay_t = data(1, :);
Range_T = delay_t>=0.5 & delay_t<=1000;
delay_T = delay_t(Range_T);
wavelength = data(1:end, 1);
E = (h*c)./(wavelength*10^-9);
Range_E = E>=1.5 & E<=2.2;
E_p = E(Range_E);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
E_fit = E_p(Range_Efit);
for i = 1:length(E_fit)
E_fit = e_fit(i);
F(i) = y(1).*exp(-(E_fit./(kB.*y(2)))) + y(3);
end
F = F(:);
end
% For loop to create required datasets for plotting
for n = 2:length(delay_T)
delta_Abs(:,1) = -1*data_new(:,1);
delta_Abs_norm(:,1) = delta_Abs(:,1)./max(abs(delta_Abs(:,1)));
delta_Abs_fit(:,1) = data_new(Range_Wfit,1);
delta_Abs_norm_fit(:,1) = delta_Abs_norm(Range_Wfit);
delta_Abs(:,n) = -1*data_new(:,n);
delta_Abs_norm(:,n) = delta_Abs(:,n)./max(abs(delta_Abs(:,n)));
delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
delta_Abs_norm_fit(:,n) = delta_Abs_norm(Range_Wfit);
% lsqcurvefit
optim_lsq = optimoptions(‘lsqcurvefit’, ‘Algorithm’, ‘levenberg-marquardt’, ‘MaxFunctionEvaluations’,10^20, ‘MaxIterations’, 10^20, ‘FunctionTolerance’,10^-20, ‘StepTolerance’, 10^-20);
lb = [0, 293, -1]; ub = [Inf, 6000, 1];
lb1 = [0, 293, -1]; ub1 = [Inf, 2000, 1];
lb2 = [0, 293, -1]; ub2 = [Inf, 1200, 1];
lb3 = [0, 293, -1]; ub3 = [Inf, 1000, 1];
lb4 = [0, 293, -1]; ub4 = [Inf, 750, 1];
y1 = [2*10^7, 1000, 0.5];
y2 = [2*10^7, 800, 0.5];
y3 = [2*10^10, 800, 0.5];
y4 = [2*10^11, 600, 0.5];
y5 = [2*10^14, 600, 0.5];
y(1,:) = lsqcurvefit(@MB, y1, E_fit, delta_Abs_norm_fit(:, 1), lb1, ub1, optim_lsq);
y(n,:) = lsqcurvefit(@MB, y(n-1, :), E_fit, delta_Abs_norm_fit(:, n), lb, ub, optim_lsq);
% plot command
plot(E_p, delta_Abs_norm)
hold on
plot(E_fit, MB(y(1,:), E_fit), ‘red’)
plot(E_fit, MB(y(n,:), E_fit), ‘red’)
xlabel(‘Probe Energy (eV)’)
ylabel(‘Normalised Delta A (a.u.)’)
legend(‘Fitted curve’)
end
carrier_T = y(:,2);
disp(carrier_T)
toc
%% Miscellaneous
% For a rectangular pic:
% drag horizontally until it just covers the letter ‘c’ in the word ‘col’
% For a square pic:
% drag horizontall until it just covers the letter ‘o’ in the word ‘Contribution’ in the next line)
% legend(‘Experimental Data’, ‘Fitted Curve’, ‘Carrier Contribution’,’Excitonic Contribution’)Hi all, I have multiple sets of data that are to be fitted by my fitting function. Since these data sets are kind of related, what I did was to use an inital guess to find the parameter values for the first dataset, and then use the parameter values as the guess for the second dataset so on and forth. However, the fitting was really bad. May I ask if there is another way around this kind of problem involving fitting of multiple datasets?
tic
%% Preparation
clear; clc
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv"); % insert file path within parenthesis
%% Preamble
% Fundamental constants
h = 4.0135667696*10^-15; % units: eV/ Hz
c = 3*10^8; % SI units
kB = 8.617333268*10^-5; % units: eV/ K
% Data
wavelength = data(1:end, 1); % contains all of the wavelengths
delay_t = data(1, 1:end); % conatains all of the delay times
E = (h*c)./(wavelength*10^-9); % contains all of the probe energies
% Range of data needed
Range_E = E>=1.5 & E<=2.2;
Range_W = wavelength >= (h*c)/(2.2*10^-9) & wavelength <= (h*c)/(1.5*10^-9);
Range_T = delay_t>=0.5 & delay_t<=1000;
E_p = E(Range_E); % probe energies for scatterplot
w_p = wavelength(Range_W);
delay_T = delay_t(Range_T);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
Range_Wfit = w_p >= (h*c)/(2.2*10^-9) & w_p <= (h*c)/(1.62*10^-9);
E_fit = E_p(Range_Efit); % probe energies for fitting
W_fit = w_p(Range_Wfit);
t_min = min(delay_T);
t_max = max(delay_T);
w_min = min(w_p);
w_max = max(w_p);
[row1, col1] = find(data == w_min);
[row2, col2] = find(data == w_max);
[row3, col3] = find(data == t_min);
[row4, col4] = find(data == t_max);
% New cleaned up data
data_new = data(row1:row2, col3:col4);% new data containing required delta A
% for n = 1:length(delay_T)
% delta_Abs(:,n) = -1*data_new(:,n);
% delta_Abs_norm(:,n) = delta_Abs(:,n)/max(delta_Abs(:,n));
% delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
% delta_Abs_norm_fit(:,n) = delta_Abs_fit(:,n)/max(delta_Abs_fit(:, n));
% % plot command
% plot(E_p, delta_Abs_norm)
% xlabel(‘Probe Energy (eV)’)
% ylabel(‘Normalised Delta A (a.u.)’)
% legend(‘Experimental Data’)
% end
% Fitting function: Maxwell-Boltzmann distribution
function F = MB(y, e_fit)
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv");
h = 4.0135667696*10^-15;
c = 3*10^8;
kB = 8.617333268*10^-5;
delay_t = data(1, :);
Range_T = delay_t>=0.5 & delay_t<=1000;
delay_T = delay_t(Range_T);
wavelength = data(1:end, 1);
E = (h*c)./(wavelength*10^-9);
Range_E = E>=1.5 & E<=2.2;
E_p = E(Range_E);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
E_fit = E_p(Range_Efit);
for i = 1:length(E_fit)
E_fit = e_fit(i);
F(i) = y(1).*exp(-(E_fit./(kB.*y(2)))) + y(3);
end
F = F(:);
end
% For loop to create required datasets for plotting
for n = 2:length(delay_T)
delta_Abs(:,1) = -1*data_new(:,1);
delta_Abs_norm(:,1) = delta_Abs(:,1)./max(abs(delta_Abs(:,1)));
delta_Abs_fit(:,1) = data_new(Range_Wfit,1);
delta_Abs_norm_fit(:,1) = delta_Abs_norm(Range_Wfit);
delta_Abs(:,n) = -1*data_new(:,n);
delta_Abs_norm(:,n) = delta_Abs(:,n)./max(abs(delta_Abs(:,n)));
delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
delta_Abs_norm_fit(:,n) = delta_Abs_norm(Range_Wfit);
% lsqcurvefit
optim_lsq = optimoptions(‘lsqcurvefit’, ‘Algorithm’, ‘levenberg-marquardt’, ‘MaxFunctionEvaluations’,10^20, ‘MaxIterations’, 10^20, ‘FunctionTolerance’,10^-20, ‘StepTolerance’, 10^-20);
lb = [0, 293, -1]; ub = [Inf, 6000, 1];
lb1 = [0, 293, -1]; ub1 = [Inf, 2000, 1];
lb2 = [0, 293, -1]; ub2 = [Inf, 1200, 1];
lb3 = [0, 293, -1]; ub3 = [Inf, 1000, 1];
lb4 = [0, 293, -1]; ub4 = [Inf, 750, 1];
y1 = [2*10^7, 1000, 0.5];
y2 = [2*10^7, 800, 0.5];
y3 = [2*10^10, 800, 0.5];
y4 = [2*10^11, 600, 0.5];
y5 = [2*10^14, 600, 0.5];
y(1,:) = lsqcurvefit(@MB, y1, E_fit, delta_Abs_norm_fit(:, 1), lb1, ub1, optim_lsq);
y(n,:) = lsqcurvefit(@MB, y(n-1, :), E_fit, delta_Abs_norm_fit(:, n), lb, ub, optim_lsq);
% plot command
plot(E_p, delta_Abs_norm)
hold on
plot(E_fit, MB(y(1,:), E_fit), ‘red’)
plot(E_fit, MB(y(n,:), E_fit), ‘red’)
xlabel(‘Probe Energy (eV)’)
ylabel(‘Normalised Delta A (a.u.)’)
legend(‘Fitted curve’)
end
carrier_T = y(:,2);
disp(carrier_T)
toc
%% Miscellaneous
% For a rectangular pic:
% drag horizontally until it just covers the letter ‘c’ in the word ‘col’
% For a square pic:
% drag horizontall until it just covers the letter ‘o’ in the word ‘Contribution’ in the next line)
% legend(‘Experimental Data’, ‘Fitted Curve’, ‘Carrier Contribution’,’Excitonic Contribution’) Hi all, I have multiple sets of data that are to be fitted by my fitting function. Since these data sets are kind of related, what I did was to use an inital guess to find the parameter values for the first dataset, and then use the parameter values as the guess for the second dataset so on and forth. However, the fitting was really bad. May I ask if there is another way around this kind of problem involving fitting of multiple datasets?
tic
%% Preparation
clear; clc
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv"); % insert file path within parenthesis
%% Preamble
% Fundamental constants
h = 4.0135667696*10^-15; % units: eV/ Hz
c = 3*10^8; % SI units
kB = 8.617333268*10^-5; % units: eV/ K
% Data
wavelength = data(1:end, 1); % contains all of the wavelengths
delay_t = data(1, 1:end); % conatains all of the delay times
E = (h*c)./(wavelength*10^-9); % contains all of the probe energies
% Range of data needed
Range_E = E>=1.5 & E<=2.2;
Range_W = wavelength >= (h*c)/(2.2*10^-9) & wavelength <= (h*c)/(1.5*10^-9);
Range_T = delay_t>=0.5 & delay_t<=1000;
E_p = E(Range_E); % probe energies for scatterplot
w_p = wavelength(Range_W);
delay_T = delay_t(Range_T);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
Range_Wfit = w_p >= (h*c)/(2.2*10^-9) & w_p <= (h*c)/(1.62*10^-9);
E_fit = E_p(Range_Efit); % probe energies for fitting
W_fit = w_p(Range_Wfit);
t_min = min(delay_T);
t_max = max(delay_T);
w_min = min(w_p);
w_max = max(w_p);
[row1, col1] = find(data == w_min);
[row2, col2] = find(data == w_max);
[row3, col3] = find(data == t_min);
[row4, col4] = find(data == t_max);
% New cleaned up data
data_new = data(row1:row2, col3:col4);% new data containing required delta A
% for n = 1:length(delay_T)
% delta_Abs(:,n) = -1*data_new(:,n);
% delta_Abs_norm(:,n) = delta_Abs(:,n)/max(delta_Abs(:,n));
% delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
% delta_Abs_norm_fit(:,n) = delta_Abs_fit(:,n)/max(delta_Abs_fit(:, n));
% % plot command
% plot(E_p, delta_Abs_norm)
% xlabel(‘Probe Energy (eV)’)
% ylabel(‘Normalised Delta A (a.u.)’)
% legend(‘Experimental Data’)
% end
% Fitting function: Maxwell-Boltzmann distribution
function F = MB(y, e_fit)
data = importdata("Experimental dataTransient AbsorptionFCPIB-293K-2.5mW-400nm-Jan072021 -ibg -bg -chirp.csv");
h = 4.0135667696*10^-15;
c = 3*10^8;
kB = 8.617333268*10^-5;
delay_t = data(1, :);
Range_T = delay_t>=0.5 & delay_t<=1000;
delay_T = delay_t(Range_T);
wavelength = data(1:end, 1);
E = (h*c)./(wavelength*10^-9);
Range_E = E>=1.5 & E<=2.2;
E_p = E(Range_E);
Range_Efit = E_p>=1.62 & E_p<=max(E_p);
E_fit = E_p(Range_Efit);
for i = 1:length(E_fit)
E_fit = e_fit(i);
F(i) = y(1).*exp(-(E_fit./(kB.*y(2)))) + y(3);
end
F = F(:);
end
% For loop to create required datasets for plotting
for n = 2:length(delay_T)
delta_Abs(:,1) = -1*data_new(:,1);
delta_Abs_norm(:,1) = delta_Abs(:,1)./max(abs(delta_Abs(:,1)));
delta_Abs_fit(:,1) = data_new(Range_Wfit,1);
delta_Abs_norm_fit(:,1) = delta_Abs_norm(Range_Wfit);
delta_Abs(:,n) = -1*data_new(:,n);
delta_Abs_norm(:,n) = delta_Abs(:,n)./max(abs(delta_Abs(:,n)));
delta_Abs_fit(:,n) = data_new(Range_Wfit,n);
delta_Abs_norm_fit(:,n) = delta_Abs_norm(Range_Wfit);
% lsqcurvefit
optim_lsq = optimoptions(‘lsqcurvefit’, ‘Algorithm’, ‘levenberg-marquardt’, ‘MaxFunctionEvaluations’,10^20, ‘MaxIterations’, 10^20, ‘FunctionTolerance’,10^-20, ‘StepTolerance’, 10^-20);
lb = [0, 293, -1]; ub = [Inf, 6000, 1];
lb1 = [0, 293, -1]; ub1 = [Inf, 2000, 1];
lb2 = [0, 293, -1]; ub2 = [Inf, 1200, 1];
lb3 = [0, 293, -1]; ub3 = [Inf, 1000, 1];
lb4 = [0, 293, -1]; ub4 = [Inf, 750, 1];
y1 = [2*10^7, 1000, 0.5];
y2 = [2*10^7, 800, 0.5];
y3 = [2*10^10, 800, 0.5];
y4 = [2*10^11, 600, 0.5];
y5 = [2*10^14, 600, 0.5];
y(1,:) = lsqcurvefit(@MB, y1, E_fit, delta_Abs_norm_fit(:, 1), lb1, ub1, optim_lsq);
y(n,:) = lsqcurvefit(@MB, y(n-1, :), E_fit, delta_Abs_norm_fit(:, n), lb, ub, optim_lsq);
% plot command
plot(E_p, delta_Abs_norm)
hold on
plot(E_fit, MB(y(1,:), E_fit), ‘red’)
plot(E_fit, MB(y(n,:), E_fit), ‘red’)
xlabel(‘Probe Energy (eV)’)
ylabel(‘Normalised Delta A (a.u.)’)
legend(‘Fitted curve’)
end
carrier_T = y(:,2);
disp(carrier_T)
toc
%% Miscellaneous
% For a rectangular pic:
% drag horizontally until it just covers the letter ‘c’ in the word ‘col’
% For a square pic:
% drag horizontall until it just covers the letter ‘o’ in the word ‘Contribution’ in the next line)
% legend(‘Experimental Data’, ‘Fitted Curve’, ‘Carrier Contribution’,’Excitonic Contribution’) lsqcurvefit MATLAB Answers — New Questions
