训练模版

Posted 2021-01-23 rongye

#!/usr/bin/env python

import tensorflow as tf



# 每个批次的大小
batch_size = 50
# 计算一共有多少个批次
#n_batch = mnist.train.num_examples // batch_size
input_num = 16
output_num = 11
def create_file(path,output_num):
    with open(path,‘r‘) as file:
        lines = file.readlines()
        count = 0
        data = []
        featuresList = []
        labelList = []
        label = []
        label3 = []
        for line in lines:
            word = line.split(" ")
            features = []
            #label3=[]
            for i in range(1, len(word)):
                if i < (len(word) - 1):
                    features.append(word[i].split(":")[1])
                else:
                    features.append(word[len(word) - 1].split(":")[1].split("
")[0])
            label.append(int(word[0]))
            count = count + 1
            #print(count)
            featuresList.append(features)
            labelList.append(label)
        for m in labelList[0]:
            label2 = []
            for k in range(output_num):
                k+=1
                #print(k,m)
                if m == k:
                    label2.append(1)
                else:
                    label2.append(0)
            label3.append(label2)
        data.append(featuresList)
        data.append(label3)
        return data[0],data[1]

data = create_file("train02.txt",output_num)
test = create_file("train02.txt",output_num)
d = tf.convert_to_tensor(data[0])#训练集
d1 = tf.convert_to_tensor(data[1])

# 初始化权值
def weight_variable(shape):
    initial = tf.truncated_normal(shape, stddev=0.1)  # 生成一个截断的正态分布
    return tf.Variable(initial)


# 初始化偏置
def bias_variable(shape):
    initial = tf.constant(0.1, shape=shape)
    return tf.Variable(initial)


# 卷积层
def conv2d(x, W):
    # x input tensor of shape `[batch, in_height, in_width, in_channels]`
    # W filter / kernel tensor of shape [filter_height, filter_width, in_channels, out_channels]
    # `strides[0] = strides[3] = 1`. strides[1]代表x方向的步长，strides[2]代表y方向的步长
    # padding: A `string` from: `"SAME", "VALID"`
    return tf.nn.conv2d(x, W, strides=[1, 1, 1, 1], padding=‘SAME‘)


# 池化层
def max_pool_2x2(x):
    # ksize [1,x,y,1]
    return tf.nn.max_pool(x, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1], padding=‘SAME‘)
# 定义两个placeholder
x = tf.placeholder(tf.float32, [None, input_num])
y = tf.placeholder(tf.float32, [None, output_num])
# 改变x的格式转为4D的向量[batch, in_height, in_width, in_channels]`
x_image = tf.reshape(x, [-1, 4, 4, 1])
# 初始化第一个卷积层的权值和偏置
W_conv1 = weight_variable([2, 2, 1, 16])  # 5*5的采样窗口，32个卷积核从1个平面抽取特征
b_conv1 = bias_variable([16])  # 每一个卷积核一个偏置值
# 把x_image和权值向量进行卷积，再加上偏置值，然后应用于relu激活函数
conv2d_1 = conv2d(x_image, W_conv1) + b_conv1
h_conv1 = tf.nn.relu(conv2d_1)
h_pool1 = max_pool_2x2(h_conv1)  # 进行max-pooling
    # 初始化第二个卷积层的权值和偏置
#W_conv2 = weight_variable([2, 2, 16, 32])  # 5*5的采样窗口，64个卷积核从32个平面抽取特征
#b_conv2 = bias_variable([32])  # 每一个卷积核一个偏置值
    # 把h_pool1和权值向量进行卷积，再加上偏置值，然后应用于relu激活函数
#conv2d_2 = conv2d(h_pool1, W_conv2) + b_conv2
#h_conv2 = tf.nn.relu(conv2d_2)
#h_pool2 = max_pool_2x2(h_conv2)  # 进行max-pooling
# 28*28的图片第一次卷积后还是28*28，第一次池化后变为14*14
# 第二次卷积后为14*14，第二次池化后变为了7*7
# 进过上面操作后得到64张7*7的平面
    # 初始化第一个全连接层的权值
W_fc1 = weight_variable([2 * 2 * 16, 512])  # 上一场有7*7*64个神经元，全连接层有1024个神经元
b_fc1 = bias_variable([512])  # 1024个节点
    # 把池化层2的输出扁平化为1维
h_pool2_flat = tf.reshape(h_pool1, [-1, 2 * 2 * 16])
    # 求第一个全连接层的输出
wx_plus_b1 = tf.matmul(h_pool2_flat, W_fc1) + b_fc1
h_fc1 = tf.nn.relu(wx_plus_b1)
    # keep_prob用来表示神经元的输出概率
keep_prob = tf.placeholder(tf.float32)
h_fc1_drop = tf.nn.dropout(h_fc1, keep_prob)
    # 初始化第二个全连接层
W_fc2 = weight_variable([512, output_num])
b_fc2 = bias_variable([output_num])
wx_plus_b2 = tf.matmul(h_fc1_drop, W_fc2) + b_fc2
        # 计算输出
prediction = tf.nn.softmax(wx_plus_b2)
# 交叉熵代价函数
cross_entropy = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits_v2(labels=y, logits=prediction))
# 使用AdamOptimizer进行优化
train_step = tf.train.AdamOptimizer(1e-4).minimize(cross_entropy)
        # 结果存放在一个布尔列表中
correct_prediction = tf.equal(tf.argmax(prediction, 1), tf.argmax(y, 1))  # argmax返回一维张量中最大的值所在的位置
        # 求准确率
accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))
saver = tf.train.Saver(max_to_keep=4)


with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    batch_xs = sess.run(d)
    batch_ys = sess.run(d1)
    n_batch = len(batch_xs)//batch_size
    print(n_batch,"n_batch")
    for epoch in range(50):
        i_item=0
        for batch in range(n_batch+1):
            begin = batch * batch_size
            if batch == n_batch:
                end = len(batch_xs)
            else:
                end = (batch + 1)*batch_size
            batch_x = batch_xs[begin:end]
            batch_y = batch_ys[begin:end]
            # print(begin)
            # print(end)
            # print(batch_y)
            sess.run(train_step, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 0.7})
        if epoch % 10 ==0:
            saver.save(sess,"model/my-model",global_step=epoch)
            acc = sess.run(accuracy, feed_dict={x: batch_xs, y: batch_ys, keep_prob: 1.0})
            print("Iter " + str(epoch) + ", Testing Accuracy= " + str(acc))