美文网首页tensorflow
tensorflow笔记(滑动平均+正则化缓解过拟合)

tensorflow笔记(滑动平均+正则化缓解过拟合)

作者: Jasmine晴天和我 | 来源:发表于2019-06-03 14:46 被阅读0次

    滑动平均

    滑动平均(影子值):记录了每个参数一段时间内过往值的平均,增加了模型的泛化性。
    针对所有参数:w和b
    (像是给参数加了影子,参数变化,影子缓慢追随)
    影子 = 衰减率影子+(1-衰减率)参数 影子初值 = 参数初值
    衰减率 = min{MOVING_AVERAGE_DECAY,\frac{1+轮数}{10+轮数}}

    ema = tf.train.ExponentialMovingAverage(衰减率MOVING_AVERAGE_DECAY(是一个超参数),当前轮数global_step)
    ema_op = ema.apply([])
    ema_op = ema.apply(tf.trainable_variables())
    每运行此句,所有待优化的参数求滑动平均
    with tf.control_dependencies([train_step,ema_op]):
        train_op = tf.no_op(name = "train")
    ema.average(参数名)  查看某参数的滑动平均值
    
    import tensorflow as tf
    
    #1. 定义变量及滑动平均类
    #定义一个32位浮点变量,初始值为0.0  这个代码就是不断更新w1参数,优化w1参数,滑动平均做了个w1的影子
    w1 = tf.Variable(0, dtype=tf.float32)
    #定义num_updates(NN的迭代轮数),初始值为0,不可被优化(训练),这个参数不训练
    global_step = tf.Variable(0, trainable=False)
    #实例化滑动平均类,给衰减率为0.99,当前轮数global_step
    MOVING_AVERAGE_DECAY = 0.99
    ema = tf.train.ExponentialMovingAverage(MOVING_AVERAGE_DECAY, global_step)
    #ema.apply后的括号里是更新列表,每次运行sess.run(ema_op)时,对更新列表中的元素求滑动平均值。
    #在实际应用中会使用tf.trainable_variables()自动将所有待训练的参数汇总为列表
    #ema_op = ema.apply([w1])
    ema_op = ema.apply(tf.trainable_variables()) #滑动平均节点
    
    #2. 查看不同迭代中变量取值的变化。
    with tf.Session() as sess:
        # 初始化
        init_op = tf.global_variables_initializer()
        sess.run(init_op)
        #用ema.average(w1)获取w1滑动平均值 (要运行多个节点,作为列表中的元素列出,写在sess.run中)
        #打印出当前参数w1和w1滑动平均值
        print ("current global_step:", sess.run(global_step))
        print ("current w1", sess.run([w1, ema.average(w1)]) )
        
        # 参数w1的值赋为1
        sess.run(tf.assign(w1, 1))
        sess.run(ema_op)
        print ("current global_step:", sess.run(global_step))
        print ("current w1", sess.run([w1, ema.average(w1)]) )
        
        # 更新global_step和w1的值,模拟出轮数为100时,参数w1变为10, 以下代码global_step保持为100,每次执行滑动平均操作,影子值会更新 
        sess.run(tf.assign(global_step, 100))  
        sess.run(tf.assign(w1, 10))
        sess.run(ema_op)
        print ("current global_step:", sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
        
        # 每次sess.run会更新一次w1的滑动平均值
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    
        sess.run(ema_op)
        print ("current global_step:" , sess.run(global_step))
        print ("current w1:", sess.run([w1, ema.average(w1)]))
    #随后每执行一次,参数w1的滑动平均都向w1逼近
    #更改MOVING_AVERAGE_DECAY 为 0.1  看影子追随速度
    #衰减率减小后,影子追随速度增加
    

    正则化缓解过拟合

    正则化在损失函数中引入模型复杂度指标,利用给W加权值,弱化了训练数据的噪声(一般不正则化b)
    loss = loss(y与y_)+REGULARIZER*loss(w)

    loss(w) = tf.contirb.layers.l1_regularizer(REGULARIZER)(w) 
    loss(w) = tf.contirb.layers.l2_regularizer(REGULARIZER)(w) 
    

    loss_{L1}(w) = \sum_{i}|w_i|
    loss_{L2}(w) = \sum_{i}|w_i^2|

    tf.add_to_collection("losses",tf.contrib.layers.l2_regularizer(regularizer)(w))
    loss = cem+tf.add_n(tf.get_collection("losses"))
    
    BATCH_SIZE = 30 
    seed = 2 
    #基于seed产生随机数
    rdm = np.random.RandomState(seed)
    print("rdm:\n",rdm)
    #随机数返回300行2列的矩阵,表示300组坐标点(x0,x1)作为输入数据集
    X = rdm.randn(300,2)
    print("X:\n",X)
    #从X这个300行2列的矩阵中取出一行,判断如果两个坐标的平方和小于2,给Y赋值1,其余赋值0
    #作为输入数据集的标签(正确答案)
    Y_ = [int(x0*x0 + x1*x1 <2) for (x0,x1) in X]
    print("Y_:\n",Y_)
    
    #遍历Y中的每个元素,1赋值'red'其余赋值'blue',这样可视化显示时人可以直观区分
    Y_c = [['red' if y else 'blue'] for y in Y_]
    #对数据集X和标签Y进行shape整理,第一个元素为-1表示,随第二个参数计算得到,第二个元素表示多少列,把X整理为n行2列,把Y整理为n行1列
    X = np.vstack(X).reshape(-1,2)
    Y_ = np.vstack(Y_).reshape(-1,1)
    print("Y_c:\n",Y_c)
    print("X:\n",X)
    print("Y_:\n",Y_)
    
    np.squeeze(Y_c) #从数组的形状中删除单维度条目,即把shape中为1的维度去掉
    
    #用plt.scatter画出数据集X各行中第0列元素和第1列元素的点即各行的(x0,x1),用各行Y_c对应的值表示颜色(c是color的缩写) 
    plt.scatter(X[:,0], X[:,1], c=np.squeeze(Y_c)) 
    plt.show()
    ![图片.png](https://img.haomeiwen.com/i11956727/0834a86d206bd429.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
    
    #定义神经网络的输入、参数和输出,定义前向传播过程 
    def get_weight(shape, regularizer): #w的shape,w的正则化权重
        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))
    
    w1 = get_weight([2,11], 0.01)   
    b1 = get_bias([11])
    y1 = tf.nn.relu(tf.matmul(x, w1)+b1)
    
    w2 = get_weight([11,1], 0.01)
    b2 = get_bias([1])
    y = tf.matmul(y1, w2)+b2 
    
    #定义损失函数
    loss_mse = tf.reduce_mean(tf.square(y-y_))
    loss_total = loss_mse + tf.add_n(tf.get_collection('losses'))
    
    
    #定义反向传播方法:不含正则化
    train_step = tf.train.AdamOptimizer(0.0001).minimize(loss_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:
                loss_mse_v = sess.run(loss_mse, feed_dict={x:X, y_:Y_})
                print("After %d steps, loss is: %f" %(i, loss_mse_v))
        #xx在-3到3之间以步长为0.01,yy在-3到3之间以步长0.01,生成二维网格坐标点
        xx, yy = np.mgrid[-3:3:.01, -3:3:.01]
        #将xx , yy拉直,并合并成一个2列的矩阵,得到一个网格坐标点的集合
        grid = np.c_[xx.ravel(), yy.ravel()]
        #将网格坐标点喂入神经网络 ,probs为输出
        probs = sess.run(y, feed_dict={x:grid})
        #probs的shape调整成xx的样子
        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()
    ![图片.png](https://img.haomeiwen.com/i11956727/b86c2a7ef1e42d17.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
    
    
    #定义反向传播方法:包含正则化
    train_step = tf.train.AdamOptimizer(0.0001).minimize(loss_total)
    
    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(loss_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()
    ![图片.png](https://img.haomeiwen.com/i11956727/42599f6d8e888f78.png?imageMogr2/auto-orient/strip%7CimageView2/2/w/1240)
    
    

    可见正则化之后的泛化效果比不加入正则的效果好。

    相关文章

      网友评论

        本文标题:tensorflow笔记(滑动平均+正则化缓解过拟合)

        本文链接:https://www.haomeiwen.com/subject/xybqxctx.html