Tensorflo函數（持續擴充）

Tensorflow函數用法（持續擴充）

1. tf.clip_by_value

tf.clip\_by\_value(v,min,max):
給定一個張量v，將張量v中地每一個元素壓縮到[min,max]的值域內。（小於min的置爲min，大於max的置爲max）。

2. tf.reduce_mean

沿着張量的指定的軸（某一維度），計算張量中元素的平均值。

#Computes the mean of elements across dimensions of a tensor
def reduce_mean(input_tensor,
	                axis=None,
	                keepdims=None,
	                name=None,
	                reduction_indices=None,
	                keep_dims=None):  

Args:

  _ input_tensor : The tensor to reduce. Should have numeric type.
  axis : The dimensions to reduce. If None (the default), reduces all dimensions. Must be in the range [-rank(input_tensor), rank(input_tensor)].
   keepdims : If true, retains reduced dimensions with length 1.
  _ name : A name for the operation (optional).
  _ reduction_indices : The old (deprecated) name for axis.
  _ keep_dims : Deprecated alias for keepdims.

Returns:
  The reduced tensor.

  x = tf.constant([1, 0, 1, 0])  
  tf.reduce_mean(x)  # 0  
  y = tf.constant([1., 0., 1., 0.])  
  tf.reduce_mean(y)  # 0.5

  x = tf.constant([[1., 1.], [2., 2.]])
  tf.reduce_mean(x)  # 1.5
  tf.reduce_mean(x, 0)  # [1.5, 1.5],以橫軸爲基準，對橫軸每一維所包含的列元素求平均。簡單來說，對行(第0維）做壓縮。
  tf.reduce_mean(x, 1)  # [1.,  2.]，壓縮列（第1維）。

3. cross_entropy(交叉熵)

Tensorflow四種交叉熵函數

• tf.nn.sigmoid_cross_entropy_with_logits
• tf.nn.softmax_cross_entropy_with_logits
• tf.nn.sparse_softmax_cross_entropy_with_logits
• tf.nn.weighted_cross_entropy_with_logits

4. tf.matmul(v1,v2):

矩陣相乘，與*不同。
*的結果爲每個元素對應位置上的乘積
matmul爲矩陣相乘

5. tf.where/tf.greater

import tensorflow as tf 
v1=tf.constant([1.,2.,3.]) 
v2=tf.constant([3.,1.,4.]) 
with tf.Session() as sess:
    Great=tf.greater(v1,v2)
    print (sess.run(Great))
    #[False  True False]
    Where=tf.where(Great,v1,v2)
    print(Where)
    #Tensor("Select:0", shape=(3,), dtype=float32)
    print(sess.run(Where))
    #[ 3.  2.  4.]

6. tf.train.exponential_decay:

#Applies exponential decay to the learning rate.
def exponential_decay(learning_rate,
                      global_step,
                      decay_steps,
                      decay_rate,
                      staircase=False,
                      name=None):  

Returns:
  The function returns the decayed learning rate. It is computed as:

  decayed_learning_rate = learning_rate *
                          decay_rate ^ (global_step / decay_steps)  

Args:

• learning_rate: A scalar float32 or float64 Tensor or a Python number. The initial learning rate.
• global_step: A scalar int32 or int64 Tensor or a Python number. Global step to use for the decay computation. Must not be negative.
• decay_steps: A scalar int32 or int64 Tensor or a Python number. Must be positive. See the decay computation above.
• decay_rate: A scalar float32 or float64 Tensor or a Python number. The decay rate. staircase: Boolean. If True decay the learning rate at discrete intervals
• name: String. Optional name of the operation. Defaults to ‘ExponentialDecay’.

7. tf.argmax:

#Returns the index with the largest value across dimensions of a tensor
def argmax(input,
           axis=None,
           name=None,
           dimension=None,
           output_type=dtypes.int64):

說明：tf\.argmax(V,1):
 V代表一個張量1表示選取最大值的操作僅在第1個維度上進行，即只在每一行選取最大值對應的下標。
實例：

import tensorflow as tf 
V=tf.constant([[1,2,3],[2,3,4]])
Max=tf.argmax(V,1)
print(Max.eval(session=tf.Session()))
#[2 2] 結果存儲的是每一行的最大值對應的下標值
Max2=tf.argmax(V,0)
print(Max2.eval(session=tf.Session()))
#[1 1 1] 結果存儲的是每一列的最大值對應的下標值