Update plotting code

src/plot_bayesian_optimization.py

@@ -0,0 +1,112 @@
+import numpy as np
+import os
+import matplotlib.pyplot as plt
+
+
+path_results = '../results'
+path_figures = '../figures'
+
+num_layerss = [2, 4, 6, 8]
+seeds = np.arange(42, 421, 42)
+num_init = 10
+num_iter = 500
+
+
+def plot(means, sems, num_layers, str_objective, show_figure, save_figure):
+    cm = plt.get_cmap('tab10')
+
+    if str_objective == 'trans':
+        str_label = r'\textrm{Transmittance}'
+        str_ylabel = r'\textrm{Maximum transmittance}'
+        color = cm(0)
+    elif str_objective == 'effec':
+        str_label = r'\textrm{Shielding effectiveness}'
+        str_ylabel = r'\textrm{Maximum shielding effectiveness (dB)}'
+        color = cm(1)
+    else:
+        raise ValueError
+
+    plt.rc('text', usetex=True)
+
+    fig = plt.figure(figsize=(8, 6))
+    ax = fig.gca()
+
+    bx = np.arange(0, means.shape[0])
+
+    ax.plot(bx, means, label=str_label, linewidth=4, linestyle='solid', color=color)
+    ax.fill_between(bx, means - sems, means + sems, alpha=0.3, color=color)
+
+    ax.set_xlabel(r'\textrm{Iteration}', fontsize=24)
+    ax.set_ylabel(str_ylabel, fontsize=24)
+
+    ax.set_xlim([0, np.max(bx)])
+    plt.tick_params(axis='both', which='major', labelsize=20)
+
+    plt.grid()
+    plt.tight_layout()
+
+    if save_figure:
+        plt.savefig(os.path.join(path_figures, f'maxima_layers_{num_layers}_{str_objective}.pdf'),
+            format='pdf', transparent=True, bbox_inches='tight')
+        plt.savefig(os.path.join(path_figures, f'maxima_layers_{num_layers}_{str_objective}.png'),
+            format='png', transparent=True, bbox_inches='tight')
+
+    if show_figure:
+        plt.show()
+
+    plt.close('all')
+
+def get_maxima(values, num_init):
+    new_values = []
+
+    for value in values:
+        new_value = [np.max(value[:num_init])]
+
+        for val in value[num_init:]:
+            if new_value[-1] < val:
+                new_value.append(val)
+            else:
+                new_value.append(new_value[-1])
+        new_values.append(new_value)
+    return np.array(new_values)
+
+def get_means_sems(values):
+    means = np.mean(values, axis=0)
+    sems = np.std(values, axis=0, ddof=1) / np.sqrt(values.shape[0])
+    return means, sems
+
+
+if __name__ == '__main__':
+    show_figure = True
+    save_figure = True
+
+    for num_layers in num_layerss:
+        transparencies_all = []
+        shielding_effectivenesses_all = []
+
+        for seed in seeds:
+            str_file = f'mobo_layers_{num_layers}_init_{num_init}_iter_{num_iter}_seed_{seed:04d}.npy'
+
+            results = np.load(os.path.join(path_results, str_file), allow_pickle=True)
+            results = results[()]
+
+            negative_transparencies = results['negative_transparencies']
+            negative_shielding_effectivenesses = results['negative_shielding_effectivenesses']
+
+            transparencies = -1.0 * negative_transparencies
+            shielding_effectivenesses = -1.0 * negative_shielding_effectivenesses
+
+            print(transparencies.shape)
+            print(shielding_effectivenesses.shape)
+
+            transparencies_all.append(transparencies)
+            shielding_effectivenesses_all.append(shielding_effectivenesses)
+
+        trans = get_maxima(transparencies_all, num_init)
+        effec = get_maxima(shielding_effectivenesses_all, num_init)
+
+        means_trans, sems_trans = get_means_sems(trans)
+        means_effec, sems_effec = get_means_sems(effec)
+
+        plot(means_trans, sems_trans, num_layers, 'trans', show_figure, save_figure)
+        plot(means_effec, sems_effec, num_layers, 'effec', show_figure, save_figure)

src/plot_pareto_frontiers.py

@@ -0,0 +1,86 @@
+import numpy as np
+import os
+import matplotlib.pyplot as plt
+
+
+path_results = '../results'
+path_figures = '../figures'
+
+num_layerss = [2, 4, 6, 8]
+seeds = np.arange(42, 421, 42)
+num_init = 10
+num_iter = 500
+
+
+def is_pareto_frontier(objs):
+    assert isinstance(objs, np.ndarray)
+    assert len(objs.shape) == 2
+    assert objs.shape[1] == 2
+
+    is_pareto = np.ones(objs.shape[0], dtype=bool)
+
+    for i, c in enumerate(objs):
+        if is_pareto[i]:
+            is_pareto[is_pareto] = np.any(objs[is_pareto] > c, axis=1)
+            is_pareto[i] = True
+    return is_pareto
+
+def plot(Y, num_layers):
+    plt.rc('text', usetex=True)
+
+    fig = plt.figure(figsize=(8, 6))
+    ax = fig.gca()
+    cm = plt.get_cmap('tab10')
+
+    pareto_frontier = Y[is_pareto_frontier(Y)]
+    indices = np.argsort(pareto_frontier[:, 0])
+    pareto_frontier = pareto_frontier[indices]
+
+    ax.plot(Y[:, 0], Y[:, 1], linestyle='none', color=cm(2), marker='.', markersize=14)
+    ax.plot(pareto_frontier[:, 0], pareto_frontier[:, 1], linestyle='solid', linewidth=4, color=cm(3))
+
+    ax.set_xlabel(r'\textrm{Transmittance}', fontsize=24)
+    ax.set_ylabel(r'\textrm{Shielding Effectiveness (dB)}', fontsize=24)
+
+    plt.tick_params(axis='both', which='major', labelsize=20)
+
+    plt.grid()
+    plt.tight_layout()
+
+    if save_figure:
+        plt.savefig(os.path.join(path_figures, f'pareto_layers_{num_layers}.pdf'), format='pdf', transparent=True, bbox_inches='tight')
+        plt.savefig(os.path.join(path_figures, f'pareto_layers_{num_layers}.png'), format='png', transparent=True, bbox_inches='tight')
+
+    if show_figure:
+        plt.show()
+
+    plt.close('all')
+
+
+if __name__ == '__main__':
+    show_figure = True
+    save_figure = False
+
+    for num_layers in num_layerss:
+        transparencies_all = []
+        shielding_effectivenesses_all = []
+
+        for seed in seeds:
+            str_file = f'mobo_layers_{num_layers}_init_{num_init}_iter_{num_iter}_seed_{seed:04d}.npy'
+
+            results = np.load(os.path.join(path_results, str_file), allow_pickle=True)
+            results = results[()]
+
+            negative_transparencies = results['negative_transparencies']
+            negative_shielding_effectivenesses = results['negative_shielding_effectivenesses']
+
+            transparencies = -1.0 * negative_transparencies
+            shielding_effectivenesses = -1.0 * negative_shielding_effectivenesses
+
+            transparencies_all += list(transparencies)
+            shielding_effectivenesses_all += list(shielding_effectivenesses)
+
+        Y = np.array([transparencies_all, shielding_effectivenesses_all]).T
+        print(Y.shape)
+
+        plot(Y, num_layers)

src/plot_structures.py

@@ -30,7 +30,7 @@ def get_color(material):
     elif material == 'W':
         str_color = 'orange'
     elif material == 'TiO2':
-        str_color = 'lawngreen'
+        str_color = 'blue'
     elif material == 'TiN':
         str_color = 'darkgreen'
     elif material == 'Al2O3':