-- 環(huán)境：win10, jupyter notebook/pycharm, python3.x, tensorflow1.3.0-gpu

上章我們學(xué)習(xí)了tensorflow怎樣運(yùn)用算法太實(shí)現(xiàn)MNIST手寫字的照片識別，這一節(jié)來建立一個(gè)簡單模型，使得識別精度更高。

Import Numpy ,Tensorflow, TFLearn, and MNIST data

import numpy as np 
import tensorflow as tf
import tflearn
import tflearn.datasets.mnist as mnist

Retrieve the training and test data

x_train, y_train, x_test, y_test = mnist.load_data(one_hot = True)

數(shù)據(jù)加載

首先講講這句話的作用，matplotlib是最著名的Python圖表繪制擴(kuò)展庫，它支持輸出多種格式的圖形圖像，并且可以使用多種GUI界面庫交互式地顯示圖表。使用%matplotlib命令可以將matplotlib的圖表直接嵌入到Notebook之中，或者使用指定的界面庫顯示圖表，它有一個(gè)參數(shù)指定matplotlib圖表的顯示方式。inline表示將圖表嵌入到Notebook中。

其次，這個(gè)問題是說沒有合理的GUI請求，所以要安裝以下任何一種都可以，ipython-qtconsole就行。
IPython提供了許多魔法命令，使得在IPython環(huán)境中的操作更加得心應(yīng)手。魔法命令都以%或者%%開頭，以%開頭的成為行命令，%%開頭的稱為單元命令。行命令只對命令所在的行有效，而單元命令則必須出現(xiàn)在單元的第一行，對整個(gè)單元的代碼進(jìn)行處理。

執(zhí)行%magic可以查看關(guān)于各個(gè)命令的說明，而在命令之后添加?可以查看該命令的詳細(xì)說明。

#Visualiziing the data
import matplotlib.pyplot as plt
%matplotlib inline

#Function for displaying a training image by it's   index in the MNIST set
def show_digit(index):
    label = y_train[index].argmax(axis=0)
    #Reshape 784 array into 28x28 image
    image = x_train[index].reshape([28,28])
    plt.title('Training data, indx:%d, Label:%d'%(index, label))
    plt.imshow(image, cmap='gray_r')
    plt.show()

#Display the first (index 0) training image
show_digit(0)

帶標(biāo)簽數(shù)據(jù)顯示

Define the meural network

def build_model():
    #This resets all parameters and variables
    tf.reset_default_graph()
    #input layer ,Hidder layer(s),and set how you   want to train the model
    #input
    net = tflearn.input_data([None, 784])
    #hidder layer
    net = tflearn.fully_connected(net,128, activation = 'ReLU')
    #net = tflearn.fully_connected(net, 64, activation = 'ReLU')
    #net = tflearn.fully_connected(net, 32, activation = 'ReLU')
    net = tflearn.fully_connected(net, 20, activation = 'ReLU')
    #Output layer
    net = tflearn.fully_connected(net, 10, activation = 'softmax')
    net = tflearn.regression(net,   optimizer='sgd',learning_rate = 0.1,\
                       loss = 'categorical_crossentropy')
    #This model assumes that your network is named 'net'
    model = tflearn.DNN(net)
    return model

此神經(jīng)網(wǎng)絡(luò)模型包含一個(gè)輸入層，兩個(gè)全連接層激活函數(shù)為ReLu，一個(gè)輸出層激活函數(shù)為softmax

Build the model

model = build_model()

training network

此處參數(shù)為調(diào)試后的最佳參數(shù)，可以修改了觀察得到結(jié)果的不同，epoch不能太小，也不能太大，對自己網(wǎng)絡(luò)訓(xùn)練效果好就得適當(dāng)

model.fit(x_train,y_train, validation_set = 0.1, show_metric = True,batch_size =10,n_epoch = 11）

訓(xùn)練網(wǎng)絡(luò)

訓(xùn)練結(jié)束

Testting

# Compare the labels that our model predicts with the actual labels

  # Find the indices of the most confident prediction for   each item. That tells us the predicted digit for that sample.
predictions =   np.array(model.predict(testX)).argmax(axis=1)

# Calculate the accuracy, which is the percentage of times the predicated labels matched the actual labels
actual = testY.argmax(axis=1)
test_accuracy = np.mean(predictions == actual, axis=0)

# Print out the result
print("Test accuracy: ", test_accuracy)

97%