深度学习与神经网络:正则化小栗子(附代码)

代码:
import tensorflow as tf
import numpy as np
import matplotlib.pyplot as plt
'''制作数据集'''
BATCH_SIZE = 30
SEED = 100
#产生随机数
rdm = np.random.RandomState(SEED)
X = rdm.randn(300,2)  #随机三百行两列的数
Y_ = [int(x0 * x0 + x1 * x1 < 2) for (x0, x1) in X]
#从X中获取,得到正确Y
Y_c = [['red' if y else 'blue']for y in Y_] #y_中标注颜色
X = np.vstack(X).reshape(-1,2)
Y_ = np.vstack(Y_).reshape(-1,1)
plt.scatter(X[:,0],X[:,1],c = np.squeeze(Y_c))#横坐标的第一二列元素
plt.show()
'''定义神经网络'''
def get_weight(shape,regularizer):
        w = tf.Variable(tf.random_normal(shape),dtype=tf.float32)
        tf.add_to_collection('losses',tf.contrib.layers.l2_regularizer(regularizer)(w))
        return w

def get_bias(shape):
    b  = tf.Variable(tf.constant(0.01,shape=shape))
    return b

x = tf.placeholder(tf.float32,shape=(None,2))
y_ = tf.placeholder(tf.float32,shape=(None,1))
#隐藏层
w1 = get_weight([2,11],0.01)
b1 = get_bias([11])
y1 = tf.nn.relu(tf.matmul(x,w1)+b1)
#输出层
w2 = get_weight([11,1],0.01)
b2 = get_bias([1])
y = tf.matmul(y1,w2)+b2
#损失函数
lose_mse  = tf.reduce_mean(tf.squeeze(y-y_))
lose_total  = lose_mse+tf.add_n(tf.get_collection('losses'))
#定义反向传播,无正则化
train_step = tf.train.AdamOptimizer(0.0001).minimize(lose_mse)

with tf.Session() as sess:
    init_op = tf.global_variables_initializer()
    sess.run(init_op)
    STEPS = 40000
    for i in range(STEPS):
        start  = (i*BATCH_SIZE)%300
        end  = start+BATCH_SIZE
        sess.run(train_step,feed_dict={x:X[start:end],y_:Y_[start:end]})
        if i%2000 ==0:
            lose_mse_v = sess.run(lose_mse,feed_dict={x:X,y_:Y_})
            print("after %d step,loss is:%f"%(i,lose_mse_v))
    xx,yy = np.mgrid[-3:3:.01,-3:3:.01]
    grid  = np.c_[xx.ravel(),yy.ravel()]
    probs  = sess.run(y,feed_dict={x:grid})
    probs = probs.reshape(xx.shape)
    print("w1:\n",sess.run(w1))
    print("b1:\n",sess.run(b1))
    print("w2:\n",sess.run(w2))
    print("b2:\n",sess.run(b1))

plt.scatter(X[:,0], X[:,1],c = np.squeeze(Y_c))
plt.contour(xx,yy,probs,levels = [.5])
plt.show()
# 定义反向传播方法：包含正则化

with tf.Session() as sess:
    init_op = tf.global_variables_initializer()
    sess.run(init_op)
    STEPS = 40000
    for i in range(STEPS):
        start = (i * BATCH_SIZE) % 300
        end = start + BATCH_SIZE
        sess.run(train_step, feed_dict={x: X[start:end], y_: Y_[start:end]})
        if i % 2000 == 0:
            loss_v = sess.run(lose_total, feed_dict={x: X, y_: Y_})
            print("After %d steps, loss is: %f" % (i, loss_v))

    xx, yy = np.mgrid[-3:3:.01, -3:3:.01]
    grid = np.c_[xx.ravel(), yy.ravel()]
    probs = sess.run(y, feed_dict={x: grid})
    probs = probs.reshape(xx.shape)
    print
    "w1:\n", sess.run(w1)
    print
    "b1:\n", sess.run(b1)
    print
    "w2:\n", sess.run(w2)
    print
    "b2:\n", sess.run(b2)

plt.scatter(X[:, 0], X[:, 1], c=np.squeeze(Y_c))
plt.contour(xx, yy, probs, levels=[.5])
plt.show()