tensorflow學習之搭建最簡單的神經網絡

這幾天在B站看莫煩的視頻,學習一波,給出視頻地址:https://www.bilibili.com/video/av16001891/?p=22網絡

先放出代碼session

#####搭建神經網絡測試
def add_layer(inputs,in_size,out_size,activation_function=None):
    Weights = tf.Variable(tf.random_normal([in_size, out_size],dtype=np.float32))
    biases = tf.Variable(tf.zeros([1,out_size])+0.1)
    Wx_plus_b = tf.matmul(inputs, Weights)+biases
    if activation_function is None:
        outputs = Wx_plus_b
    else:
        outputs = activation_function(Wx_plus_b)
    return outputs

x_data = np.linspace(-1,1,300)[:, np.newaxis]
noise = np.random.normal(0,0.05,x_data.shape)
y_data = np.square(x_data)-0.5+noise

xs = tf.placeholder(tf.float32,[None,1])
ys = tf.placeholder(tf.float32,[None,1])
l1 = add_layer(xs,1,10,activation_function=tf.nn.relu)

prediction = add_layer(l1,10,1,activation_function=None)
loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),
                                   reduction_indices=[1]))
train_step = tf.train.GradientDescentOptimizer(0.1).minimize(loss)

init = tf.global_variables_initializer()
with tf.Session() as sess:
    sess.run(init)
    for i in range(1000):
        sess.run(train_step,feed_dict={xs:x_data,ys:y_data})
        if i% 50 ==0:
            print(sess.run(loss,feed_dict={xs:x_data,ys:y_data}))
#####

  首先,在add_layer函數中,參數有inputs,in_size,out_size,activation_function=Nonedom

其中inupts是輸入,in_size是輸入維度,out_size是輸出維度, activation_function是激活函數,ide

Weights是權重,維度是(in_size*out_size);函數

bias是偏置,維度是(1*out_size);學習

Wx_plus_b的維度和out_size相同;測試

  x_data = np.linspace(-1,1,300)[:, np.newaxis]這步操做,表示生成-1到1之間均勻分佈的300個數,而後轉換維度,變成(300,1);noise和y_data的維度均和優化

x_data相同;spa

  xs = tf.placeholder(tf.float32,[None,1])和ys = tf.placeholder(tf.float32,[None,1])表示生成xs和ys變量的佔位符,維度是(None,1),不知道有多少行,但只要1列;code

  l1 = add_layer(xs,1,10,activation_function=tf.nn.relu)表示xs是inputs,in_size是1,out_size是10,激活函數是relu;添加了一層神經網絡

  prediction = add_layer(l1,10,1,activation_function=None)表示輸入是l1,in_size是10,out_size是1,沒有激活函數

  接下去是計算損失,loss = tf.reduce_mean(tf.reduce_sum(tf.square(ys - prediction),reduction_indices=[1]))

  以後一步是用梯度降低來優化損失函數;

解釋一下爲何不直接在add_layer函數中使用x_data:x_data是ndarray格式,Weights是Variable格式,不能直接相乘,因此要在session會話中用字典格式傳入x_data和y_data,  也就是sess.run(train_step,feed_dict={xs:x_data,ys:y_data})

相關文章
相關標籤/搜索