###### 加载所需的模块
from sklearn import datasets
from matplotlib import pyplot as plt
import numpy as np
from sklearn.preprocessing import StandardScaler
from sklearn.model_selection import train_test_split
from sklearn.neural_network import MLPClassifier
from matplotlib.colors import ListedColormap
from mpl_toolkits.mplot3d import Axes3D
from matplotlib.animation import FuncAnimation
from matplotlib import colors
import time
%matplotlib notebook

#几种激活函数
def relu(x):
    return np.where(x <= 0,0,x)

def sigmoid(x):
    return 1/(1+np.exp(-x))

def tanh(x):
    return (np.exp(x) - np.exp(-x)) / (np.exp(x) + np.exp(-x))

#生成样本
X,y= datasets.make_circles(n_samples = 2000, factor=0.3, noise=.1)

#划分训练集和测试集
X, X_test, y, y_test = train_test_split(X, y, test_size=0.33, random_state=42)

#构建特征空间，这一步是不是应该在划分训练集和测试集之前啊
c,r = np.mgrid[[slice(X.min() - .5,X.max() + .5,50j)]*2]
#这里为什么要加flat呀 加不加的结果都是一样的啊 迭代器
p = np.c_[c.flat,r.flat]
#p是坐标空间

#归一化
%matplotlib inline
ss = StandardScaler().fit(X)
X = ss.transform(X)
p = ss.transform(p)
X_test = ss.transform(X_test)


#可视化
fig = plt.figure(figsize = (9,3))
cm_bright = ListedColormap(['#f32b1a','#0c6dea'])
plt.subplot(121)
m1 = plt.scatter(*X.T,c = y,cmap = cm_bright,edgecolors='white')
plt.title('train samples')
plt.axis('equal')
plt.subplot(122)
m2 = plt.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white');
plt.title('test samples')
plt.axis('equal')
ax = fig.get_axes()
plt.colorbar(ax = ax);
plt.show();

#分类
loss,train,test = [], [], []
MLP = MLPClassifier(hidden_layer_sizes=(4,2),max_iter = 20, activation='relu',
                    warm_start = True,learning_rate_init=0.008,)
start = time.time()
#训练了500次，epoch
for i in range(500):
    MLP.fit(X,y)
    loss.append(MLP.loss_)
    train.append(MLP.score(X,y))
    test.append(MLP.score(X_test,y_test))
#记录训练时间
    end = time.time()
print (end-start)
...;

C:\Users\nxjhz\Anaconda3\lib\site-packages\sklearn\neural_network\multilayer_perceptron.py:564: ConvergenceWarning: Stochastic Optimizer: Maximum iterations (20) reached and the optimization hasn't converged yet.
  % self.max_iter, ConvergenceWarning)

2.210728168487549

#记录权重 和偏置量
W,B = MLP.coefs_ , MLP.intercepts_
W
B
z = MLP.predict(p)
...;

#数据训练情况
%matplotlib inline
fig, ((ax1,ax2),(ax3,ax4)) = plt.subplots(2,2, figsize=(6, 6))
ax1.scatter(*p.T,c = z,cmap = cm_bright)
ax1.scatter(*X.T,c = y,cmap = cm_bright,edgecolors='white')
ax1.set_title('train score:%.5f'%train[-1])

ax2.scatter(*p.T,c = z,cmap = cm_bright)
ax2.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')
ax2.set_title('test score:%.5f'%test[-1])

ax3.plot(loss)
ax3.set_title('Loss')

ax4.plot(train,c = 'r',label = 'train')
ax4.plot(test,c = 'b',label = 'test')
ax4.set_title('train-test score');
ax4.legend();
plt.show();

%matplotlib inline
#第一隐层
layer_one0 = (W[0].T) @ (p.T)
b = np.array([B[0]]).T
layer_one0 = layer_one0 + b

#非线性化操作
layer_one = relu(layer_one0)

vmin = layer_one0.min()
vmax = layer_one0.max()
norm = colors.Normalize(vmin=vmin, vmax=vmax)

cmap_name = 'bwr'
fig, ((ax1,ax2,ax3,ax4),(ax5,ax6,ax7,ax8)) = plt.subplots(2,4, figsize=(20, 8))
plt.rcParams['font.sans-serif']=['SimHei'] #用来正常显示中文标签
plt.rcParams['axes.unicode_minus']=False #用来正常显示负号

#作图 8张图
c1 = ax1.scatter(*p.T,c = layer_one0[0],cmap = cmap_name, norm = norm)
ax1.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax1.set_title('第二层的第一个节点')


c2 =ax2.scatter(*p.T,c = layer_one0[1],cmap = cmap_name, norm = norm)
ax2.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax2.set_title('第二层的第二个节点')

ax3.scatter(*p.T,c = layer_one0[2],cmap = cmap_name, norm = norm)
ax3.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax3.set_title('第二层的第三个节点')

ax4.scatter(*p.T,c = layer_one0[3],cmap = cmap_name, norm = norm)
ax4.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax4.set_title('第二层的第四个节点')

ax5.scatter(*p.T,c = layer_one[0],cmap = cmap_name, norm = norm)
ax5.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax5.set_title('非线性：第二层的第一个节点')

c6 = ax6.scatter(*p.T,c = layer_one[1],cmap = cmap_name, norm = norm)
ax6.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax6.set_title('非线性：第二层的第二个节点')

ax7.scatter(*p.T,c = layer_one[2],cmap = cmap_name, norm = norm)
ax7.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax7.set_title('非线性：第二层的第三个节点')

ax8.scatter(*p.T,c = layer_one[3],cmap = cmap_name, norm = norm)
ax8.scatter(*X_test.T,c = y_test,cmap = cmap_name,edgecolors='white')
ax8.set_title('非线性：第二层的第四个节点')

plt.colorbar(c2,ax=[ax1,ax2,ax3,ax4,ax5,ax6,ax7,ax8])
...;

#第一隐层的胞腔划分 二维空间
%matplotlib inline
plt.figure(figsize=(6,6))
layer_one_split = np.where(layer_one<=0,0,1)
layer_one_split_result=layer_one_split[0]*1+layer_one_split[1]*2+layer_one_split[2]*4+layer_one_split[3]*8
c2 =plt.scatter(*p.T,c = layer_one_split_result,cmap = 'tab20')
# plt.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')

<Figure size 432x432 with 0 Axes>

% matplotlib notebook
#前三个cell动态图（画后三个cell的时候*x_.T[:3]改成*x_.T[1:]，
#layer_one[0],layer_one[1],layer_one[2]改成layer_one[1],layer_one[2],layer_one[3]
layer_one_split = np.where(layer_one<=0,0,1)
layer_one_split_result=layer_one_split[0]*1+layer_one_split[1]*2+layer_one_split[2]*4

x_ = relu(X @ W[0] + B[0])

fig,(ax1,ax2) = plt.subplots(1,2,figsize = (9,5),subplot_kw = {'projection':'3d'})
# ax1.scatter(layer_one[0],layer_one[1],layer_one[2],c = layer_one_split_result,cmap = 'tab10')
ax1.scatter(*x_.T[:3],c = y,cmap = cm_bright,edgecolors='white')
ax1.elev,ax1.azim = 15, -60

ax2.scatter(layer_one[0],layer_one[1],layer_one[2],c = layer_one_split_result,cmap = 'tab10')
ax2.scatter(*x_.T[:3],c = y,cmap = cm_bright,edgecolors='white')
def update(i):
    ax2.azim = i * 10
    ax2.elev = i * 10
    return ax2
ani = FuncAnimation(fig, update, 36,  interval=500);

#下载安装imagemagick，添加环境变量
ani.save('d3.gif', writer='imagemagick')
plt.show();
...;

Traceback (most recent call last):
  File "C:\Users\nxjhz\Anaconda3\lib\site-packages\matplotlib\cbook\__init__.py", line 216, in process
    func(*args, **kwargs)
  File "C:\Users\nxjhz\Anaconda3\lib\site-packages\matplotlib\animation.py", line 1465, in _stop
    self.event_source.remove_callback(self._loop_delay)
AttributeError: 'NoneType' object has no attribute 'remove_callback'

% matplotlib inline
#第二隐层
#作图 共4张
layer_two = (W[1].T) @ layer_one
b = np.array([B[1]]).T
layer_two = layer_two + b


fig, ((ax1,ax2),(ax3,ax4)) = plt.subplots(2,2, figsize=(9, 6))
plt.rcParams['font.sans-serif'] = ['SimHei']
vmin = -5
vmax = 5
norm = colors.Normalize(vmin=vmin, vmax=vmax)

ax1.scatter(*p.T,c = layer_two[0],cmap = cmap_name, norm = norm)
ax1.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')
ax1.set_title('第三层的第一个节点')

ax2.scatter(*p.T,c = layer_two[1],cmap = cmap_name, norm = norm)
ax2.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')
ax2.set_title('第三层的第二个节点')
#非线性
layer_two = relu(layer_two)

ax3.scatter(*p.T,c = layer_two[0],cmap = cmap_name, norm = norm)
ax3.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')
ax3.set_title('非线性：第三层的第一个节点')

c2 = ax4.scatter(*p.T,c = layer_two[1],cmap = cmap_name, norm = norm)
ax4.scatter(*X_test.T,c = y_test,cmap = cm_bright,edgecolors='white')
ax4.set_title('非线性：第二层的第二个节点')

plt.colorbar(c2,ax=[ax1,ax2,ax3,ax4])
...;

%matplotlib inline
#第二隐层（第三层）的二维平面图
x1_ = relu(X @ W[0] + B[0])
x2_ = relu(x1_ @ W[1] + B[1])
p1_ = relu(p @ W[0] + B[0])
p2_ = relu(p1_ @ W[1] + B[1])
fig, (ax1,ax2,ax3) = plt.subplots(1,3, figsize=(12, 6))
layer_two_split = np.where(layer_two<=0,0,1)
layer_two_split_result=layer_two_split[0]*1+layer_two_split[1]*2
c2 =ax1.scatter(*p2_.T,c = layer_two_split_result,cmap = cmap_name)
ax2.scatter(*x2_.T,c = y, cmap = cmap_name, edgecolors='white')
ax3.scatter(*p2_.T,c = layer_two_split_result,cmap = cmap_name)
ax3.scatter(*x2_.T,c = y, cmap = cmap_name, edgecolors='white')
...;

#输出层
#作图 分别为非线性变换前和非线性变换后
layer_two = relu(layer_two)
layer_three = (W[2].T) @ (layer_two)
layer_three = layer_three + B[2].T

x1_ = relu(X @ W[0] + B[0])
x2_ = relu(x1_ @ W[1] + B[1])
x3_ = x2_ @ W[2] + B[2]

p1_ = relu(p @ W[0] + B[0])
p2_ = relu(p1_ @ W[1] + B[1])
p3_ = p2_ @ W[2] + B[2]

plt.rcParams['font.sans-serif'] = ['SimHei']
fig, ((ax1,ax2,ax3),(ax4,ax5,ax6)) = plt.subplots(2,3, figsize=(15, 9))


vmin = p3_.min()
vmax = p3_.max()
vmin = -vmax if vmax > -vmin else vmin
vmax = -vmin if vmax < -vmin else vmax
norm = colors.Normalize(vmin=vmin, vmax=vmax)

c2 = ax1.scatter(*p3_.T,*p3_.T,c = layer_three[0],cmap = cmap_name, norm = norm)
ax2.scatter(*x3_.T,*x3_.T,c = y,cmap = cmap_name,edgecolors='white')
ax3.scatter(*p3_.T,*p3_.T,c = layer_three[0],cmap = cmap_name, norm = norm)
ax3.scatter(*x3_.T,*x3_.T,c = y,cmap = cmap_name,edgecolors='white')
ax1.set_title('输出层特征空间')
ax2.set_title('输出层数据点')
ax3.set_title('合并')

layer_three = relu(layer_three)
x3_ = relu(x2_ @ W[2] + B[2])
p3_ = relu(p2_ @ W[2] + B[2])


ax4.scatter(*p3_.T,*p3_.T,c = layer_three[0],cmap = cmap_name, norm = norm)
ax5.scatter(*x3_.T,*x3_.T,c = y,cmap =cmap_name,edgecolors='white')
ax6.scatter(*p3_.T,*p3_.T,c = layer_three[0],cmap = cmap_name, norm = norm)
ax6.scatter(*x3_.T,*x3_.T,c = y,cmap =cmap_name,edgecolors='white')
ax4.set_title('非线性：输出层特征空间')
ax5.set_title('非线性：输出层数据点')
ax6.set_title('非线性：合并')

plt.colorbar(c2,ax=[ax1,ax2,ax3,ax4,ax5,ax6])
...;