tensorflow套路

1数据读取

2准备好placeholder

X = tf.placeholder(tf.float32, [batch_size, 784], name='X_placeholder') 
Y_true = tf.placeholder(tf.int32, [batch_size, 10], name='Y_placeholder')

3参数初始化
w = tf.Variable(tf.random_normal(shape=[784, 10], stddev=0.01), name='weights')
b = tf.Variable(tf.zeros([1, 10]), name="bias")

4计算预测值

5计算损失值

#均方误差
loss = tf.reduce_mean(tf.square(y_true - y_predict))

# 求平均交叉熵损失
loss = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(labels=y_true, logits=y_predict))

6优化器
train_op = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

###############################计算准确率

equal_list = tf.equal(tf.argmax(y_true, 1), tf.argmax(y_predict, 1))
#参数1代表对每一行的数据求最大值的索引
二维数据的索引为0 ，1：
# equal_list  None个样本   [1, 0, 1, 0, 1, 1,..........]
accuracy = tf.reduce_mean(tf.cast(equal_list, tf.float32))

###########################多层神经网络

# 网络参数
n_hidden_1 = 256 # 第1个隐层
n_hidden_2 = 256 # 第2个隐层
n_input = 784 # MNIST 数据输入(28*28*1=784)
n_classes = 10 # MNIST 总共10个手写数字类别

weights = {
    'h1': tf.Variable(tf.random_normal([n_input, n_hidden_1]), name='W1'),
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2]), name='W2'),
    'out': tf.Variable(tf.random_normal([n_hidden_2, n_classes]), name='W')
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1]), name='b1'),
    'b2': tf.Variable(tf.random_normal([n_hidden_2]), name='b2'),
    'out': tf.Variable(tf.random_normal([n_classes]), name='bias')
}

def multilayer_perceptron(x, weights, biases):
    # 第1个隐层，使用relu激活函数
    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'], name='fc_1')
    layer_1 = tf.nn.relu(layer_1, name='relu_1')
    # 第2个隐层，使用relu激活函数
    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'], name='fc_2')
    layer_2 = tf.nn.relu(layer_2, name='relu_2')
    # 输出层
    out_layer = tf.add(tf.matmul(layer_2, weights['out']), biases['out'], name='fc_3')
    return out_layer

拿到预测类别score
pred = multilayer_perceptron(X, weights, biases)