plot_local_error_circuit.m: New script to illustrate the circuit example

scripts/alpha_investigation.m

@@ -67,3 +67,5 @@
 alpha_without_dominant_poles = SzegoSymModel_without_dominant_poles.param(2)*1e4
 
 alpha_with_dominant_poles = SzegoSymModel_with_dominant_poles.param(2)*1e4
+
+% for the selected value of alpha for the hybrid method, see the script plot_local_error_circuit.m

scripts/plot_local_error_circuit.m

@@ -0,0 +1,92 @@
+%% Plot illustration of different approximations and associated pointwise error for Circuit example
+clf; close all; clear;
+
+%% Function to approximate (Circuit with/without dominant poles)
+% f = @(omega) fun_Circuit(omega, 1000, 1,0.2, true); xmin=1; xmax=2.5; fun='Circuit'; 
+f = @(omega) fun_Circuit(omega, 1000, 1,0.2, false); xmin=1; xmax=2.5; fun='Circuit'; 
+
+%% Training points:
+n_training_points = 50;
+xi = linspace(xmin, xmax, n_training_points)';
+yi = f(xi);
+
+%% Test points
+n_cv = 501;
+%x_cv = unique(sort([linspace(xmin,xmax,n_cv)',xi]));
+x_cv = linspace(xmin,xmax,n_cv)';
+y_cv = f(x_cv);
+
+
+%% Set up Export Data
+export_data = zeros(n_cv, 10); header = {};
+export_data(:, 1) = 1e4 * x_cv; header{1} = 'x';
+export_data(:, 2) = abs(y_cv); header{2} = 'y_ref_abs';
+
+
+%% Illustrate function:
+plot_cplxfun(x_cv,y_cv, 'Reference', 'k-')
+plot_cplxfun(xi, yi, 'Data for Fitting', 'kx')
+
+%% AAA
+[approxAAA, polesAAA] = aaa(yi,xi);
+plot_cplxfun(x_cv, approxAAA(x_cv), 'AAA', '--');
+plot_cplxfun(x_cv, approxAAA(x_cv)-y_cv, 'Error AAA', 'b-',  [], [], [], 100)
+figure(1000)
+semilogy(x_cv, abs(approxAAA(x_cv)-y_cv), 'b-', 'DisplayName', 'AAA')
+hold on
+export_data(:, 3) = abs(approxAAA(x_cv)); header{3} = 'AAA';
+export_data(:, 4) = abs(approxAAA(x_cv)-y_cv); header{4} = 'Error_AAA';
+
+[RMSE_AAA, max_errorAAA] = compute_approx_error(approxAAA(x_cv), y_cv, 'AAA');
+
+%% Vector Fitting (using complex starting poles):
+[approxVF, polesVF] = VectorFitting(xi,yi);
+plot_cplxfun(x_cv, approxVF(x_cv), 'VectorFitting Approx', '--');
+plot_cplxfun(x_cv, approxVF(x_cv)-y_cv, 'Error VF', 'r--',  [], [], [], 100)
+figure(1000)
+semilogy(x_cv, abs(approxVF(x_cv)-y_cv), 'r--', 'DisplayName', 'VF')
+[RMSE_VF, max_errorVF] = compute_approx_error(approxVF(x_cv), y_cv, 'VF');
+export_data(:, 5) = abs(approxVF(x_cv)); header{5} = 'VF';
+export_data(:, 6) = abs(approxVF(x_cv)-y_cv); header{6} = 'Error_VF';
+
+
+%% Szegö
+opts=init_opts();
+[Mean, Var, Var_real, Var_imag, crit_opt, model] = CplxGPapprox('Szego', xi, yi, x_cv, opts);%, [-5; -1], [5;5]);%,[],[],[],0,[1, 1e-3]);
+alpha_Szego = model.param(2)*1e4
+plot_cplxfun(x_cv, Mean, ['Szego'], 'b:');
+plot_cplxfun(x_cv, Mean-y_cv, 'Error Szegö', 'g-',  [], [], [], 100)
+figure(1000)
+semilogy(x_cv, abs(Mean-y_cv), 'g-', 'DisplayName', 'Szegö')
+[RMSE_Szego, max_errorSzego] = compute_approx_error(Mean, y_cv, 'Szego');
+export_data(:, 7) = abs(Mean); header{7} = 'Szego';
+export_data(:, 8) = abs(Mean-y_cv); header{8} = 'Error_Szego';
+
+%% Adaptive Approximation
+adap_opts = init_adap_opts();
+%opts.verbose=true;
+adap_opts.verbose=true;
+opts.n_restart=5;
+[Mean,Var, Var_real, Var_imag,crit,adap_model,res, data] = AdapApprox('Szego', xi, yi, x_cv,opts, adap_opts);
+alpha_hybrid = adap_model.param(2)*1e4
+plot_cplxfun(x_cv, Mean, 'Adap Approx', '--');
+plot_cplxfun(x_cv, Mean-y_cv, 'Error Adap', 'k-',  [], [], [], 100)
+figure(1000)
+semilogy(x_cv, abs(Mean-y_cv), 'k-', 'DisplayName', 'Adap Approx')
+title('Error (Magnitude)')
+legend;
+export_data(:, 9) = abs(Mean); header{9} = 'Adap';
+export_data(:, 10) = abs(Mean-y_cv); header{10} = 'Error_Adap';
+
+
+[RMSE_Adap, max_errorAdap] = compute_approx_error(Mean, y_cv, 'Adap');
+for i = 1:length(data)
+    compute_approx_error(data{i}.Mean, y_cv, ['Model with ' num2str(i-1) ' poles']);
+end
+
+for i = 1:length(data)
+    data{i}.model.param(2)*1e4
+end
+
+export_csv('results/illustration_approx_circuit.csv', export_data, header);
+

scripts/run_all.m

@@ -6,3 +6,4 @@
 run_ConvStudies;
 alpha_investigation;
 test_higher_kmax.m
+plot_local_error_circuit.m