differential-evolution/create_animation.py at master · jannisko/differential-evolution · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
from differential_evolution import DifferentialEvolution
from differential_evolution.losses import create_huber
from differential_evolution.models import create_quadratic
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
from matplotlib import animation, cm
import sys

def animate_models(
    output_gif=False,
    output_mp4=True,
    frame_num=300,
    fps=30
):

    x = [1,2,3,4,5,6,7,8]
    y = [1,4,9,16,25,36,49,64]

    a = tf.Variable(0.0)
    b = tf.Variable(0.0)
    c = tf.Variable(0.0)

    model = create_quadratic(x, a, b, c)

    loss = create_huber(model, y)

    fig = plt.figure(figsize=(4.5, 3.375), dpi=100)
    ax = fig.gca()

    plot_model = create_quadratic(np.linspace(0,10,100), a, b, c)

    opt = DifferentialEvolution(loss, [a,b,c])

    def animate(i):
        opt.next_generation()
        plt.cla()
        for point in opt.current_population:
            a.assign(point[0][0])
            b.assign(point[1][0])
            c.assign(point[2][0])
            ax.plot(np.linspace(0,10,100), plot_model(), zorder=-1)
        ax.scatter(x,y, zorder=1, marker='^', c='000000')
        sys.stdout.write("\r%d%%" % ((i/frame_num)*100))
        sys.stdout.flush()
        ax.set_ylim((0, 70))
        ax.set_xlim((0, 10))
        plt.xlabel('x', fontsize=8)
        plt.ylabel('x²', fontsize=8)
        plt.text(5, 72,
            f'Epoch: {i:03}/{frame_num}; Best loss: {opt.get_best_point()[1]:08.5f}',
            fontsize=7
        )
        #plt.title('An example of a quadratic regression using differential evolution')
        for t in ax.xaxis.get_major_ticks(): t.label.set_fontsize(7)
        for t in ax.yaxis.get_major_ticks(): t.label.set_fontsize(7)
        return fig,


    def init():
        return fig,

    anim = animation.FuncAnimation(fig, animate, init_func=init,
                                frames=frame_num, interval=100, blit=True)


    if(output_gif):
        anim.save('animations/quadratic_models.gif', fps=fps, writer='imagemagick')

    if(output_mp4):
        Writer = animation.writers['ffmpeg']
        writer = Writer(fps=fps, bitrate=3000)
        anim.save('animations/quadratic_models.mp4', writer=writer)


from mpl_toolkits.mplot3d import Axes3D
from differential_evolution import ackley

def animate_ackley(
    output_gif=False,
    output_mp4=True,
    frame_num=180,
    fps=30,
    frames_per_update=15,
    generations_per_update=5
):

    # figsize was chosen to fit github's README size
    fig = plt.figure(figsize=(4.5, 3.375), dpi=100)
    #fig = plt.figure()
    ax = fig.gca(projection='3d')
    # remove background grid, fill and axis
    ax.grid(False)
    ax.xaxis.pane.fill = ax.yaxis.pane.fill = ax.zaxis.pane.fill = False
    plt.axis('off')
    # tighter fit to window
    plt.tight_layout()

    # create surface values
    x = np.arange(-5, 5, 0.05)
    y = np.arange(-5, 5, 0.05)
    xx, yy = np.meshgrid(x, y)
    z = np.empty(xx.shape, dtype=np.float)
    for i, (px, py) in enumerate(zip(xx,yy)):
        z[i] = ackley.ackley_function(px, py)

    a = tf.Variable(0.0)
    b = tf.Variable(0.0)

    # wrapper function so ackley can be called everytime a and b change
    loss = lambda : ackley.ackley_function(a,b)

    opt = DifferentialEvolution(loss, [a,b], boundaries=5)

    def animate(i):

        ax.view_init(elev=30.0, azim=i)


        if(i % frames_per_update == 0):
            for _ in range(generations_per_update):
                opt.next_generation()
            plt.cla()
            ax.plot_surface(xx, yy, z, cmap=cm.gnuplot,
                    linewidth=0, antialiased=False, alpha=0.3)
            population = opt.get_current_population()
            ax.scatter(
                population.T[0],
                population.T[1],
                ackley.ackley_function(population.T[0], population.T[1]),
                marker='o', s=40, c='#000000', alpha=1
            )
            ax.text2D(
                0.5, 0.97,
                f'Epoch: {i:03}/{frame_num}; Best loss: {opt.get_best_point()[1]:08.5f}',
                fontsize=7,
                transform=ax.transAxes
            )

        sys.stdout.write("\r%d%%" % ((i/frame_num)*100))
        sys.stdout.flush()

        ax.set_ylim((-5, 5))
        ax.set_xlim((-5, 5))
        ax.set_zlim((0, 15))

        ax.set_xlabel('x', fontsize=8)
        ax.set_ylabel('y', fontsize=8)
        ax.set_zlabel('ackley(x,y)', fontsize=8)

        for t in ax.xaxis.get_major_ticks(): t.label.set_fontsize(7)
        for t in ax.yaxis.get_major_ticks(): t.label.set_fontsize(7)
        for t in ax.zaxis.get_major_ticks(): t.label.set_fontsize(7)

        return fig,

    def init():
        return fig,

    anim = animation.FuncAnimation(fig, animate, init_func=init,
                                frames=frame_num, interval=100, blit=True)


    if(output_gif):
        anim.save('animations/ackley_optimization.gif', fps=fps, writer='imagemagick')

    if(output_mp4):
        Writer = animation.writers['ffmpeg']
        writer = Writer(fps=fps, bitrate=3000)
        anim.save('animations/ackley_optimization.mp4', writer=writer)

if __name__ == "__main__":
    animate_models(output_gif=True, output_mp4=False)
    animate_ackley(output_gif=True, output_mp4=False)