美文网首页
The Computational Graph

The Computational Graph

作者: harukou_ou | 来源:发表于2017-12-28 22:33 被阅读68次

    Tensor

    3            # a rank 0 tensor; a scalar with shape []
    [1., 2., 3.]   # a rank 1 tensor; a vector with shape [3]
    [[1., 2., 3.], [4., 5., 6.]]   # a rank 2 tensor; a matrix with shape [2, 3]
    [[[1., 2., 3.]], [[7., 8., 9.]]]   # a rank 3 tensor with shape [2, 1, 3]
    

    Rank: 类似矩阵中的秩,一般来说,看开头的中括号的数目决定其个数
    Shape:对张量进行描述,形式为[x,y,z,...]。其中括号内的个数与rank个数相同。eg:以上述最后一个为例,[2,1,3]中2代表逗号“ , ”前后一共有2个rank=2,每个都为1×3的矩阵。


    Importing TensorFlow

    import tensorflow as tf 
    

    Constant

    One type of node is a constant. it takes no inputs, and it outputs a value it stores internally. We can create two floating point Tensors node1 and node2 as follows:

    node1 = tf.constant(3.0, dtype=tf.float32)
    node2 = tf.constant(4.0) # also tf.float32 implicitly
    print(node1)
    print(node2)
    

    print statement produces:

    Tensor("Const:0", shape=(), dtype=float32) 
    Tensor("Const_1:0", shape=(), dtype=float32)
    

    they are nodes that, until when evaluated, would produce 3.0 and 4.0, respectively. The following code creates a Session object and then invokes its run method to run enough of the computational graph to evaluate node1 and node2. By running the computational graph in a session as follows:

    sess = tf.Session() # session is important!
    print(sess.run([node1, node2]))
    
    

    we see the expected values of 3.0 and 4.0:

    [3.0, 4.0]
    

    Add

    Add our two constant nodes and produce a new graph as follows:

    node3 = tf.add(node1, node2)  #also node3=node1+node2
    print("node3:", node3)
    print("sess.run(node3):", sess.run(node3))
    

    print statements produce:

    node3: Tensor("Add:0", shape=(), dtype=float32)
    sess.run(node3): 7.0
    

    TensorBoard displays a picture of the computation graph like below:
    [站外图片上传中...(image-d30106-1514471593029)]


    Placeholder

    parameterized to accept external inputs, known as placeholders. A placeholder is a promise to provide a value later.

    a = tf.placeholder(tf.float32)
    b = tf.placeholder(tf.float32)
    adder_node = a + b  # + likes tf.add(a, b)
    print(sess.run(adder_node, {a: 3, b: 4.5}))
    print(sess.run(adder_node, {a: [1, 3], b: [2, 4]}))
    

    sess的run方法中第二个参数为字典

    7.5
    [ 3.  7.]
    

    In TensorBoard, the graph looks like this:
    [站外图片上传中...(image-d53209-1514471593029)]
    more complex by adding another operation:

    add_and_triple = adder_node * 3.
    print(sess.run(add_and_triple, {a: 3, b: 4.5}))
    

    resulting:
    22.5
    [站外图片上传中...(image-e26c5a-1514471593029)]


    Variable

    In machine learning we can take arbitrary inputs, such as the one above. Variables allow us to add trainable parameters to a graph. They are constructed with a type and initial value:

    W = tf.Variable([.3], dtype=tf.float32)
    b = tf.Variable([-.3], dtype=tf.float32)
    x = tf.placeholder(tf.float32)
    linear_model = W*x + b
    

    注意:Variable 需要初始化,方法:

    init = tf.global_variables_initializer() # 全局变量?
    sess.run(init)
    

    Since x is a placeholder, we can evaluate linear_model for several values of x simultaneously as follows:

    print(sess.run(linear_model, {x: [1, 2, 3, 4]}))
    

    to pruduce:

    [ 0.          0.30000001  0.60000002  0.90000004]
    

    Loss function

    To evaluate the model on training data, we need a y placeholder to provide the desired values, and we need to write a loss function.

    y = tf.placeholder(tf.float32)
    squared_deltas = tf.square(linear_model - y) # (y`-y)^2
    loss = tf.reduce_sum(squared_deltas) #    求和
    print(sess.run(loss, {x: [1, 2, 3, 4], y: [0, -1, -2, -3]}))
    

    得到:

    23.66  #   23.66>>0 显然W,b不符合要求
    

    假设我们找到了W=-1.0,b=1.0:

    fixW = tf.assign(W, [-1.]) 
    fixb = tf.assign(b, [1.]) 
    sess.run([fixW, fixb]) 
    print(sess.run(loss, {x: [1, 2, 3, 4], y: [0, -1, -2, -3]}))
    

    结果 loss=0,得到需要的W,b


    参考:

    https://www.tensorflow.org/get_started/get_started

    相关文章

      网友评论

          本文标题:The Computational Graph

          本文链接:https://www.haomeiwen.com/subject/gccbgxtx.html