Add files via upload

Author: Evolutionary-Intelligence

docs/pypop7_benchmarks_lso/single-core/compare_efficiency_for_de.py

@@ -0,0 +1,98 @@
+"""This script plots the *median* convergence curves for
+    various DE w.r.t. actual runtime (to be needed).
+
+    Chinese: 该绘图脚本被段琦琦的博士论文（哈工大与南科大联合培养）所使用。
+"""
+import os
+import sys
+import pickle5 as pickle
+
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+from pypop7.optimizers.core import optimizer
+sys.modules['optimizer'] = optimizer
+
+
+def read_pickle(a, f, i):
+    folder = './docs/pypop7_benchmarks_lso/single-core'
+    file_name = a + '/Algo-' + a + '_Func-' + f +\
+        '_Dim-2000_Exp-' + i + '.pickle'
+    with open(os.path.join(folder, file_name), 'rb') as handle:
+        return pickle.load(handle)
+
+
+if __name__ == '__main__':
+    font_size = 11
+    sns.set_theme(style='darkgrid')
+    plt.rcParams['font.family'] = 'Times New Roman'
+    plt.rcParams['font.size'] = '11'
+
+    n_trials = 4  # number of trials (independent experiments)
+    algos = ['DEAPDE', 'CDE', 'TDE', 'CODE', 'JADE', 'SHADE']
+    max_runtime, fitness_threshold = 3600.0 * 3 - 10.0 * 60, 1e-10
+    # funcs = ['cigar', 'cigar_discus', 'ackley', 'bohachevsky',
+    #          'different_powers', 'discus', 'griewank', 'levy_montalvo',
+    #          'ellipsoid', 'rosenbrock', 'michalewicz', 'rastrigin',
+    #          'schwefel12', 'schwefel221', 'salomon', 'scaled_rastrigin',
+    #          'step', 'schaffer', 'sphere', 'skew_rastrigin']
+    funcs = ['cigar', 'ackley', 'bohachevsky',
+             'discus', 'ellipsoid', 'rosenbrock',
+             'rastrigin', 'schwefel12', 'schwefel221']
+    fig = plt.figure(figsize=(9, 11))
+    no_of_rows, no_of_cols = 3, 3 # 5, 4
+    axs = fig.subplots(no_of_rows, no_of_cols)
+    for k, f in enumerate(funcs):
+        time, fitness = [], []
+        for i in range(len(algos)):
+            time.append([])
+            fitness.append([])
+            for _ in range(n_trials):
+                time[i].append([])
+                fitness[i].append([])
+        for i in range(n_trials):
+            for j, a in enumerate(algos):
+                results = read_pickle(a, f, str(i + 1))
+                time[j][i] = results['fitness'][:, 0] *\
+                    results['runtime'] /\
+                        results['n_function_evaluations']
+                y = results['fitness'][:, 1]
+                if f == 'michalewicz':
+                    y += 600  # to plot log-scale y-axis
+                fitness[j][i] = y
+        top_order = []
+        for j, a in enumerate(algos):
+            run, fit, r_f = [], [], []
+            for i in range(len(time[j])):
+                run.append(time[j][i][-1] if time[j][i][-1] <= max_runtime else max_runtime)
+                fit.append(fitness[j][i][-1] if fitness[j][i][-1] >= fitness_threshold else fitness_threshold)
+                r_f.append([run[i], fit[i], i])
+            r_f.sort(key=lambda x: (x[0], x[1]))
+            order = r_f[int(n_trials / 2)][2]  # for median (but non-standard)
+            top_order.append(order)
+        m, n = divmod(k, no_of_cols)
+        axs[m][n].set_yscale('log')
+        for j, a in enumerate(algos):
+            axs[m][n].plot(time[j][top_order[j]],
+                           fitness[j][top_order[j]],
+                           label=a)
+        axs[m][n].set_title(f, fontsize=font_size)  # , fontweight='bold'
+        x_label = axs[m][n].get_xticklabels()
+        [xl.set_fontsize(font_size) for xl in x_label]
+        # [xl.set_fontweight('bold') for xl in x_label]
+        y_label = axs[m][n].get_yticklabels()
+        [yl.set_fontsize(font_size) for yl in y_label]
+        # [yl.set_fontweight('bold') for yl in y_label]
+
+    lines, labels = axs[-1][-1].get_legend_handles_labels()
+    leg = fig.legend(lines, labels, loc='center', ncol=6,
+                     fontsize=font_size,
+                     bbox_to_anchor=(0.51, 0.93))
+    # for text in leg.get_texts():
+    #     text.set_fontweight('bold')
+    fig.text(0.05, 0.5, 'Fitness (Minimized)', va='center',
+             rotation='vertical', fontsize=font_size)  # 'xx-large'
+    fig.text(0.5, 0.05, 'Runtime (Seconds)', va='center',
+             ha='center', fontsize=font_size)  # 'xx-large'
+    plt.savefig('DE.png', dpi=700, bbox_inches='tight')
+    plt.show()

docs/pypop7_benchmarks_lso/single-core/compare_efficiency_for_eda.py

@@ -0,0 +1,99 @@
+"""This script plots the *median* convergence curves for
+    various EDA/CEM w.r.t. actual runtime (to be needed).
+
+    Chinese: 该绘图脚本被段琦琦的博士论文（哈工大与南科大联合培养）所使用。
+"""
+import os
+import sys
+import pickle5 as pickle
+
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+from pypop7.optimizers.core import optimizer
+sys.modules['optimizer'] = optimizer
+
+
+def read_pickle(a, f, i):
+    folder = './docs/pypop7_benchmarks_lso/single-core'
+    file_name = a + '/Algo-' + a + '_Func-' + f +\
+        '_Dim-2000_Exp-' + i + '.pickle'
+    with open(os.path.join(folder, file_name), 'rb') as handle:
+        return pickle.load(handle)
+
+
+if __name__ == '__main__':
+    font_size = 11
+    sns.set_theme(style='darkgrid')
+    plt.rcParams['font.family'] = 'Times New Roman'
+    plt.rcParams['font.size'] = '11'
+
+    n_trials = 4  # number of trials (independent experiments)
+    algos = ['DEAPEDA', 'EMNA', 'AEMNA', 'UMDA', 'RPEDA',
+             'DCEM', 'DSCEM', 'MRAS', 'SCEM']
+    max_runtime, fitness_threshold = 3600.0 * 3 - 10.0 * 60, 1e-10
+    # funcs = ['cigar', 'cigar_discus', 'ackley', 'bohachevsky',
+    #          'different_powers', 'discus', 'griewank', 'levy_montalvo',
+    #          'ellipsoid', 'rosenbrock', 'michalewicz', 'rastrigin',
+    #          'schwefel12', 'schwefel221', 'salomon', 'scaled_rastrigin',
+    #          'step', 'schaffer', 'sphere', 'skew_rastrigin']
+    funcs = ['cigar', 'ackley', 'bohachevsky',
+             'discus', 'ellipsoid', 'rosenbrock',
+             'rastrigin', 'schwefel12', 'schwefel221']
+    fig = plt.figure(figsize=(9, 11))
+    no_of_rows, no_of_cols = 3, 3 # 5, 4
+    axs = fig.subplots(no_of_rows, no_of_cols)
+    for k, f in enumerate(funcs):
+        time, fitness = [], []
+        for i in range(len(algos)):
+            time.append([])
+            fitness.append([])
+            for _ in range(n_trials):
+                time[i].append([])
+                fitness[i].append([])
+        for i in range(n_trials):
+            for j, a in enumerate(algos):
+                results = read_pickle(a, f, str(i + 1))
+                time[j][i] = results['fitness'][:, 0] *\
+                    results['runtime'] /\
+                        results['n_function_evaluations']
+                y = results['fitness'][:, 1]
+                if f == 'michalewicz':
+                    y += 600  # to plot log-scale y-axis
+                fitness[j][i] = y
+        top_order = []
+        for j, a in enumerate(algos):
+            run, fit, r_f = [], [], []
+            for i in range(len(time[j])):
+                run.append(time[j][i][-1] if time[j][i][-1] <= max_runtime else max_runtime)
+                fit.append(fitness[j][i][-1] if fitness[j][i][-1] >= fitness_threshold else fitness_threshold)
+                r_f.append([run[i], fit[i], i])
+            r_f.sort(key=lambda x: (x[0], x[1]))
+            order = r_f[int(n_trials / 2)][2]  # for median (but non-standard)
+            top_order.append(order)
+        m, n = divmod(k, no_of_cols)
+        axs[m][n].set_yscale('log')
+        for j, a in enumerate(algos):
+            axs[m][n].plot(time[j][top_order[j]],
+                           fitness[j][top_order[j]],
+                           label=a)
+        axs[m][n].set_title(f, fontsize=font_size)  # , fontweight='bold'
+        x_label = axs[m][n].get_xticklabels()
+        [xl.set_fontsize(font_size) for xl in x_label]
+        # [xl.set_fontweight('bold') for xl in x_label]
+        y_label = axs[m][n].get_yticklabels()
+        [yl.set_fontsize(font_size) for yl in y_label]
+        # [yl.set_fontweight('bold') for yl in y_label]
+
+    lines, labels = axs[-1][-1].get_legend_handles_labels()
+    leg = fig.legend(lines, labels, loc='center', ncol=5,
+                     fontsize=font_size,
+                     bbox_to_anchor=(0.51, 0.93))
+    # for text in leg.get_texts():
+    #     text.set_fontweight('bold')
+    fig.text(0.05, 0.5, 'Fitness (Minimized)', va='center',
+             rotation='vertical', fontsize=font_size)  # 'xx-large'
+    fig.text(0.5, 0.05, 'Runtime (Seconds)', va='center',
+             ha='center', fontsize=font_size)  # 'xx-large'
+    plt.savefig('EDA.png', dpi=700, bbox_inches='tight')
+    plt.show()

docs/pypop7_benchmarks_lso/single-core/compare_efficiency_for_es.py

@@ -0,0 +1,103 @@
+"""This script plots the *median* convergence curves for
+    various ES w.r.t. actual runtime (to be needed).
+
+    Chinese: 该绘图脚本被段琦琦的博士论文（哈工大与南科大联合培养）所使用。
+"""
+import os
+import sys
+import pickle5 as pickle
+
+import seaborn as sns
+import matplotlib.pyplot as plt
+
+from pypop7.optimizers.core import optimizer
+sys.modules['optimizer'] = optimizer
+
+
+def read_pickle(a, f, i):
+    folder = './docs/pypop7_benchmarks_lso/single-core'
+    file_name = a + '/Algo-' + a + '_Func-' + f +\
+        '_Dim-2000_Exp-' + i + '.pickle'
+    with open(os.path.join(folder, file_name), 'rb') as handle:
+        return pickle.load(handle)
+
+
+if __name__ == '__main__':
+    font_size = 11
+    sns.set_theme(style='darkgrid')
+    plt.rcParams['font.family'] = 'Times New Roman'
+    plt.rcParams['font.size'] = '11'
+
+    n_trials = 4  # number of trials (independent experiments)
+    algos = ['DEAPCMAES', 'MMES', 'FCMAES', 'LMMAES', 'LMCMA',
+             'RMES', 'R1ES', 'VKDCMA', 'VDCMA', 'CCMAES2016',
+             'OPOA2015', 'OPOA2010', 'CCMAES2009', 'OPOC2009', 'OPOC2006',
+             'SEPCMAES', 'DDCMA', 'MAES', 'FMAES', 'SAES',
+             'CSAES', 'RES', 'R1NES']
+    max_runtime, fitness_threshold = 3600.0 * 3 - 10.0 * 60, 1e-10
+    # funcs = ['cigar', 'cigar_discus', 'ackley', 'bohachevsky',
+    #          'different_powers', 'discus', 'griewank', 'levy_montalvo',
+    #          'ellipsoid', 'rosenbrock', 'michalewicz', 'rastrigin',
+    #          'schwefel12', 'schwefel221', 'salomon', 'scaled_rastrigin',
+    #          'step', 'schaffer', 'sphere', 'skew_rastrigin']
+    funcs = ['cigar', 'ackley', 'bohachevsky',
+             'discus', 'ellipsoid', 'rosenbrock',
+             'rastrigin', 'schwefel12', 'schwefel221']
+
+    fig = plt.figure(figsize=(9, 11))
+    no_of_rows, no_of_cols = 3, 3 # 5, 4
+    axs = fig.subplots(no_of_rows, no_of_cols)
+    for k, f in enumerate(funcs):
+        time, fitness = [], []
+        for i in range(len(algos)):
+            time.append([])
+            fitness.append([])
+            for _ in range(n_trials):
+                time[i].append([])
+                fitness[i].append([])
+        for i in range(n_trials):
+            for j, a in enumerate(algos):
+                results = read_pickle(a, f, str(i + 1))
+                time[j][i] = results['fitness'][:, 0] *\
+                    results['runtime'] /\
+                        results['n_function_evaluations']
+                y = results['fitness'][:, 1]
+                if f == 'michalewicz':
+                    y += 600  # to plot log-scale y-axis
+                fitness[j][i] = y
+        top_order = []
+        for j, a in enumerate(algos):
+            run, fit, r_f = [], [], []
+            for i in range(len(time[j])):
+                run.append(time[j][i][-1] if time[j][i][-1] <= max_runtime else max_runtime)
+                fit.append(fitness[j][i][-1] if fitness[j][i][-1] >= fitness_threshold else fitness_threshold)
+                r_f.append([run[i], fit[i], i])
+            r_f.sort(key=lambda x: (x[0], x[1]))
+            order = r_f[int(n_trials / 2)][2]  # for median (but non-standard)
+            top_order.append(order)
+        m, n = divmod(k, no_of_cols)
+        axs[m][n].set_yscale('log')
+        for j, a in enumerate(algos):
+            axs[m][n].plot(time[j][top_order[j]],
+                           fitness[j][top_order[j]],
+                           label=a)
+        axs[m][n].set_title(f, fontsize=font_size)  # , fontweight='bold'
+        x_label = axs[m][n].get_xticklabels()
+        [xl.set_fontsize(font_size) for xl in x_label]
+        # [xl.set_fontweight('bold') for xl in x_label]
+        y_label = axs[m][n].get_yticklabels()
+        [yl.set_fontsize(font_size) for yl in y_label]
+        # [yl.set_fontweight('bold') for yl in y_label]
+
+    lines, labels = axs[-1][-1].get_legend_handles_labels()
+    leg = fig.legend(lines, labels, loc='center', ncol=5,
+                     fontsize=font_size,
+                     bbox_to_anchor=(0.51, 0.95))
+    # for text in leg.get_texts():
+    #     text.set_fontweight('bold')
+    fig.text(0.05, 0.5, 'Fitness (Minimized)', va='center',
+             rotation='vertical', fontsize=font_size)  # 'xx-large'
+    fig.text(0.5, 0.05, 'Runtime (Seconds)', va='center',
+             ha='center', fontsize=font_size)  # 'xx-large'
+    plt.savefig('ES.png', dpi=700, bbox_inches='tight')
+    plt.show()