本實(shí)驗(yàn)通過建立一個(gè)含有兩個(gè)隱含層的BP神經(jīng)網(wǎng)絡(luò)，擬合具有二次函數(shù)非線性關(guān)系的方程，并通過可視化展現(xiàn)學(xué)習(xí)到的擬合曲線，同時(shí)隨機(jī)給定輸入值，輸出預(yù)測值，最后給出一些關(guān)鍵的提示。

源代碼如下：

# -*- coding: utf-8 -*- import tensorflow as tf import numpy as np import matplotlib.pyplot as pltplotdata = { "batchsize":[], "loss":[] } def moving_average(a, w=11):if len(a) < w: return a[:] return [val if idx < w else sum(a[(idx-w):idx])/w for idx, val in enumerate(a)]#生成模擬數(shù)據(jù)，二次函數(shù)關(guān)系 train_X = np.linspace(-1, 1, 100)[:, np.newaxis] train_Y = train_X*train_X + 5 * train_X + np.random.randn(*train_X.shape) * 0.3 #子圖1顯示模擬數(shù)據(jù)點(diǎn) plt.figure(12) plt.subplot(221) plt.plot(train_X, train_Y, 'ro', label='Original data') plt.legend()# 創(chuàng)建模型 # 占位符 X = tf.placeholder("float",[None,1]) Y = tf.placeholder("float",[None,1]) # 模型參數(shù) W1 = tf.Variable(tf.random_normal([1,10]), name="weight1") b1 = tf.Variable(tf.zeros([1,10]), name="bias1") W2 = tf.Variable(tf.random_normal([10,6]), name="weight2") b2 = tf.Variable(tf.zeros([1,6]), name="bias2") W3 = tf.Variable(tf.random_normal([6,1]), name="weight3") b3 = tf.Variable(tf.zeros([1]), name="bias3")# 前向結(jié)構(gòu) z1 = tf.matmul(X, W1) + b1 z2 = tf.nn.relu(z1) z3 = tf.matmul(z2, W2) + b2 z4 = tf.nn.relu(z3) z5 = tf.matmul(z4, W3) + b3#反向優(yōu)化 cost =tf.reduce_mean( tf.square(Y - z5)) learning_rate = 0.01 optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost) #Gradient descent# 初始化變量 init = tf.global_variables_initializer() # 訓(xùn)練參數(shù) training_epochs = 5000 display_step = 2# 啟動(dòng)session with tf.Session() as sess:sess.run(init)for epoch in range(training_epochs+1):sess.run(optimizer, feed_dict={X: train_X, Y: train_Y})#顯示訓(xùn)練中的詳細(xì)信息if epoch % display_step == 0:loss = sess.run(cost, feed_dict={X: train_X, Y:train_Y})print ("Epoch:", epoch, "cost=", loss)if not (loss == "NA" ):plotdata["batchsize"].append(epoch)plotdata["loss"].append(loss)print (" Finish")#圖形顯示plt.subplot(222) plt.plot(train_X, train_Y, 'ro', label='Original data')plt.plot(train_X, sess.run(z5, feed_dict={X: train_X}), label='Fitted line')plt.legend() plotdata["avgloss"] = moving_average(plotdata["loss"])plt.subplot(212)plt.plot(plotdata["batchsize"], plotdata["avgloss"], 'b--')plt.xlabel('Minibatch number')plt.ylabel('Loss')plt.title('Minibatch run vs Training loss') plt.show()#預(yù)測結(jié)果a=[[0.2],[0.3]]print ("x=[[0.2],[0.3]]，z5=", sess.run(z5, feed_dict={X: a}))

運(yùn)行結(jié)果如下：

結(jié)果實(shí)在是太棒了，把這個(gè)關(guān)系擬合的非常好。在上述的例子中，需要進(jìn)一步說明如下內(nèi)容：

• 輸入節(jié)點(diǎn)可以通過字典類型定義，而后通過字典的方法訪問

input = {'X': tf.placeholder("float",[None,1]),'Y': tf.placeholder("float",[None,1]) }
sess.run(optimizer, feed_dict={input['X']: train_X, input['Y']: train_Y})

直接定義輸入節(jié)點(diǎn)的方法是不推薦使用的。

• 變量也可以通過字典類型定義，例如上述代碼可以改為：

parameter = {'W1': tf.Variable(tf.random_normal([1,10]), name="weight1"),'b1': tf.Variable(tf.zeros([1,10]), name="bias1"),'W2': tf.Variable(tf.random_normal([10,6]), name="weight2"),'b2': tf.Variable(tf.zeros([1,6]), name="bias2"),'W3': tf.Variable(tf.random_normal([6,1]), name="weight3"),'b3': tf.Variable(tf.zeros([1]), name="bias3") } z1 = tf.matmul(X, parameter['W1']) +parameter['b1']

在上述代碼中練習(xí)保存/載入模型，代碼如下：

# -*- coding: utf-8 -*- import tensorflow as tf import numpy as np import matplotlib.pyplot as pltplotdata = { "batchsize":[], "loss":[] } def moving_average(a, w=11):if len(a) < w: return a[:] return [val if idx < w else sum(a[(idx-w):idx])/w for idx, val in enumerate(a)]#生成模擬數(shù)據(jù)，二次函數(shù)關(guān)系 train_X = np.linspace(-1, 1, 100)[:, np.newaxis] train_Y = train_X*train_X + 5 * train_X + np.random.randn(*train_X.shape) * 0.3 #子圖1顯示模擬數(shù)據(jù)點(diǎn) plt.figure(12) plt.subplot(221) plt.plot(train_X, train_Y, 'ro', label='Original data') plt.legend()# 創(chuàng)建模型 # 字典型占位符 input = {'X':tf.placeholder("float",[None,1]),'Y':tf.placeholder("float",[None,1])} # X = tf.placeholder("float",[None,1]) # Y = tf.placeholder("float",[None,1]) # 模型參數(shù) parameter = {'W1':tf.Variable(tf.random_normal([1,10]), name="weight1"), 'b1':tf.Variable(tf.zeros([1,10]), name="bias1"), 'W2':tf.Variable(tf.random_normal([10,6]), name="weight2"),'b2':tf.Variable(tf.zeros([1,6]), name="bias2"), 'W3':tf.Variable(tf.random_normal([6,1]), name="weight3"), 'b3':tf.Variable(tf.zeros([1]), name="bias3")} # W1 = tf.Variable(tf.random_normal([1,10]), name="weight1") # b1 = tf.Variable(tf.zeros([1,10]), name="bias1") # W2 = tf.Variable(tf.random_normal([10,6]), name="weight2") # b2 = tf.Variable(tf.zeros([1,6]), name="bias2") # W3 = tf.Variable(tf.random_normal([6,1]), name="weight3") # b3 = tf.Variable(tf.zeros([1]), name="bias3")# 前向結(jié)構(gòu) z1 = tf.matmul(input['X'], parameter['W1']) + parameter['b1'] z2 = tf.nn.relu(z1) z3 = tf.matmul(z2, parameter['W2']) + parameter['b2'] z4 = tf.nn.relu(z3) z5 = tf.matmul(z4, parameter['W3']) + parameter['b3']#反向優(yōu)化 cost =tf.reduce_mean( tf.square(input['Y'] - z5)) learning_rate = 0.01 optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost) #Gradient descent# 初始化變量 init = tf.global_variables_initializer() # 訓(xùn)練參數(shù) training_epochs = 5000 display_step = 2 # 生成saver saver = tf.train.Saver() savedir = "model/"# 啟動(dòng)session with tf.Session() as sess:sess.run(init)for epoch in range(training_epochs+1):sess.run(optimizer, feed_dict={input['X']: train_X, input['Y']: train_Y})#顯示訓(xùn)練中的詳細(xì)信息if epoch % display_step == 0:loss = sess.run(cost, feed_dict={input['X']: train_X, input['Y']:train_Y})print ("Epoch:", epoch, "cost=", loss)if not (loss == "NA" ):plotdata["batchsize"].append(epoch)plotdata["loss"].append(loss)print (" Finish")#保存模型saver.save(sess, savedir+"mymodel.cpkt")#圖形顯示plt.subplot(222) plt.plot(train_X, train_Y, 'ro', label='Original data')plt.plot(train_X, sess.run(z5, feed_dict={input['X']: train_X}), label='Fitted line')plt.legend() plotdata["avgloss"] = moving_average(plotdata["loss"])plt.subplot(212)plt.plot(plotdata["batchsize"], plotdata["avgloss"], 'b--')plt.xlabel('Minibatch number')plt.ylabel('Loss')plt.title('Minibatch run vs Training loss') plt.show()#預(yù)測結(jié)果 #在另外一個(gè)session里面載入保存的模型,再測試 a=[[0.2],[0.3]] with tf.Session() as sess2:#sess2.run(tf.global_variables_initializer())可有可無,因?yàn)橄旅鎟estore會(huì)載入?yún)?shù)，相當(dāng)于本次調(diào)用的初始化 saver.restore(sess2, "model/mymodel.cpkt")print ("x=[[0.2],[0.3]]，z5=", sess2.run(z5, feed_dict={input['X']: a}))

?生成如下目錄：

上述代碼模型的載入沒有利用到檢查點(diǎn)文件，顯得不夠智能，還需用戶去查找指定某一模型，那在很多算法項(xiàng)目中是不需要用戶去找的，而可以通過檢查點(diǎn)找到保存的模型。例如：

# -*- coding: utf-8 -*- import tensorflow as tf import numpy as np import matplotlib.pyplot as pltplotdata = { "batchsize":[], "loss":[] } def moving_average(a, w=11):if len(a) < w: return a[:] return [val if idx < w else sum(a[(idx-w):idx])/w for idx, val in enumerate(a)]#生成模擬數(shù)據(jù)，二次函數(shù)關(guān)系 train_X = np.linspace(-1, 1, 100)[:, np.newaxis] train_Y = train_X*train_X + 5 * train_X + np.random.randn(*train_X.shape) * 0.3 #子圖1顯示模擬數(shù)據(jù)點(diǎn) plt.figure(12) plt.subplot(221) plt.plot(train_X, train_Y, 'ro', label='Original data') plt.legend()# 創(chuàng)建模型 # 字典型占位符 input = {'X':tf.placeholder("float",[None,1]),'Y':tf.placeholder("float",[None,1])} # X = tf.placeholder("float",[None,1]) # Y = tf.placeholder("float",[None,1]) # 模型參數(shù) parameter = {'W1':tf.Variable(tf.random_normal([1,10]), name="weight1"), 'b1':tf.Variable(tf.zeros([1,10]), name="bias1"), 'W2':tf.Variable(tf.random_normal([10,6]), name="weight2"),'b2':tf.Variable(tf.zeros([1,6]), name="bias2"), 'W3':tf.Variable(tf.random_normal([6,1]), name="weight3"), 'b3':tf.Variable(tf.zeros([1]), name="bias3")} # W1 = tf.Variable(tf.random_normal([1,10]), name="weight1") # b1 = tf.Variable(tf.zeros([1,10]), name="bias1") # W2 = tf.Variable(tf.random_normal([10,6]), name="weight2") # b2 = tf.Variable(tf.zeros([1,6]), name="bias2") # W3 = tf.Variable(tf.random_normal([6,1]), name="weight3") # b3 = tf.Variable(tf.zeros([1]), name="bias3")# 前向結(jié)構(gòu) z1 = tf.matmul(input['X'], parameter['W1']) + parameter['b1'] z2 = tf.nn.relu(z1) z3 = tf.matmul(z2, parameter['W2']) + parameter['b2'] z4 = tf.nn.relu(z3) z5 = tf.matmul(z4, parameter['W3']) + parameter['b3']#反向優(yōu)化 cost =tf.reduce_mean( tf.square(input['Y'] - z5)) learning_rate = 0.01 optimizer = tf.train.GradientDescentOptimizer(learning_rate).minimize(cost) #Gradient descent# 初始化變量 init = tf.global_variables_initializer() # 訓(xùn)練參數(shù) training_epochs = 5000 display_step = 2 # 生成saver saver = tf.train.Saver(max_to_keep=1) savedir = "model/"# 啟動(dòng)session with tf.Session() as sess:sess.run(init)for epoch in range(training_epochs+1):sess.run(optimizer, feed_dict={input['X']: train_X, input['Y']: train_Y})saver.save(sess, savedir+"mymodel.cpkt",global_step=epoch)#顯示訓(xùn)練中的詳細(xì)信息if epoch % display_step == 0:loss = sess.run(cost, feed_dict={input['X']: train_X, input['Y']:train_Y})print ("Epoch:", epoch, "cost=", loss)if not (loss == "NA" ):plotdata["batchsize"].append(epoch)plotdata["loss"].append(loss)print (" Finish")#圖形顯示plt.subplot(222) plt.plot(train_X, train_Y, 'ro', label='Original data')plt.plot(train_X, sess.run(z5, feed_dict={input['X']: train_X}), label='Fitted line')plt.legend() plotdata["avgloss"] = moving_average(plotdata["loss"])plt.subplot(212)plt.plot(plotdata["batchsize"], plotdata["avgloss"], 'b--')plt.xlabel('Minibatch number')plt.ylabel('Loss')plt.title('Minibatch run vs Training loss') plt.show()#預(yù)測結(jié)果 #在另外一個(gè)session里面載入保存的模型,再測試 a=[[0.2],[0.3]] load=5000 with tf.Session() as sess2:#sess2.run(tf.global_variables_initializer())可有可無,因?yàn)橄旅鎟estore會(huì)載入?yún)?shù)，相當(dāng)于本次調(diào)用的初始化 #saver.restore(sess2, "model/mymodel.cpkt")saver.restore(sess2, "model/mymodel.cpkt-" + str(load))print ("x=[[0.2],[0.3]]，z5=", sess2.run(z5, feed_dict={input['X']: a})) #通過檢查點(diǎn)文件載入保存的模型 with tf.Session() as sess3:ckpt = tf.train.get_checkpoint_state(savedir)if ckpt and ckpt.model_checkpoint_path:saver.restore(sess3, ckpt.model_checkpoint_path)print ("x=[[0.2],[0.3]],z5=", sess3.run(z5, feed_dict={input['X']: a})) #通過檢查點(diǎn)文件載入最新保存的模型 with tf.Session() as sess4:ckpt = tf.train.latest_checkpoint(savedir)if ckpt!=None:saver.restore(sess4, ckpt) print ("x=[[0.2],[0.3]],z5=", sess4.run(z5, feed_dict={input['X']: a}))

而通常情況下，上述兩種通過檢查點(diǎn)載入模型參數(shù)的結(jié)果是一樣的，主要是因?yàn)椴还苡脩舯４媪硕嗌賯€(gè)模型文件，都會(huì)被記錄在唯一一個(gè)檢查點(diǎn)文件中，這個(gè)指定保存模型個(gè)數(shù)的參數(shù)就是max_to_keep，例如：

saver = tf.train.Saver(max_to_keep=3)

而檢查點(diǎn)都會(huì)默認(rèn)用最新的模型載入，忽略了之前的模型，因此上述兩個(gè)檢查點(diǎn)載入了同一個(gè)模型，自然最后輸出的測試結(jié)果是一致的。保存的三個(gè)模型如圖：

?

接下來，為什么上面的變量，需要給它對應(yīng)的操作起個(gè)名字，而且是不一樣的名字呢？像weight1、bias1等等。大家都知道，名字這個(gè)東西太重要了，通過它可以訪問我們想訪問的變量，也就可以對其進(jìn)行一些操作。例如：

• 顯示模型的內(nèi)容

不同版本的函數(shù)會(huì)有些區(qū)別，本文試驗(yàn)的版本是1.7.0，代碼例如：

# -*- coding: utf-8 -*- import tensorflow as tf from tensorflow.python.tools import inspect_checkpoint as chkp#顯示全部變量的名字和值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-5000", all_tensor_names='', tensor_name='', all_tensors=True) #顯示指定名字變量的值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-5000", all_tensor_names='', tensor_name='weight1', all_tensors=False) chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-5000", all_tensor_names='', tensor_name='bias1', all_tensors=False)

運(yùn)行結(jié)果如下圖：

相反如果對不同變量的操作用了同一個(gè)name，系統(tǒng)將會(huì)自動(dòng)對同名稱操作排序，例如：

# -*- coding: utf-8 -*- import tensorflow as tf from tensorflow.python.tools import inspect_checkpoint as chkp#顯示全部變量的名字和值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='', all_tensors=True) #顯示指定名字變量的值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='weight', all_tensors=False) chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='bias', all_tensors=False)

結(jié)果為：

需要注意的是因?yàn)閷λ型淖兞颗判蛑?#xff0c;真正的變量名已經(jīng)變了，所以，當(dāng)指定查看某一個(gè)變量的值時(shí)，其實(shí)輸出的是第一個(gè)變量的值，因?yàn)樗拿Q還保留著不變。另外，也可以通過變量的name屬性查看其操作名。

• 按名字保存變量

可以通過指定名稱來保存變量；注意如果名字如果搞混了，名稱所對應(yīng)的值也就搞混了，比如：

#只保存這兩個(gè)變量，并且這兩個(gè)被搞混了 saver = tf.train.Saver({'weight': parameter['b2'], 'bias':parameter['W1']})# -*- coding: utf-8 -*- import tensorflow as tf from tensorflow.python.tools import inspect_checkpoint as chkp#顯示全部變量的名字和值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='', all_tensors=True) #顯示指定名字變量的值 chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='weight', all_tensors=False) chkp.print_tensors_in_checkpoint_file("model/mymodel.cpkt-50", all_tensor_names='', tensor_name='bias', all_tensors=False)

此時(shí)的結(jié)果是：

這樣，模型按照我們的想法保存了參數(shù)，注意不能搞混變量和其對應(yīng)的名字。

?

轉(zhuǎn)載于:https://www.cnblogs.com/cvtoEyes/p/9063291.html

總結(jié)

以上是生活随笔為你收集整理的tensorflow神经网络拟合非线性函数与操作指南的全部內(nèi)容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網(wǎng)站內(nèi)容還不錯(cuò)，歡迎將生活随笔推薦給好友。