目录
tf.contrib.layers.xavier_initializer
tf.nn.l2_normalize的使用
tf.nn.l2_normalize(x, dim, epsilon=1e-12, name=None)
上式:
x为输入的向量;
dim为l2范化的维数,dim取值为0或0或1;
epsilon的范化的最小值边界;
只能理解为:dim=0时,按列归一,dim=1时按行归一
代码:
结果:
tf.nn.embedding_lookup:
https://blog.csdn.net/laolu1573/article/details/77170407
tf.nn.embedding_lookup()就是根据input_ids中的id,寻找embeddings中的第id行。比如input_ids=[1,3,5],则找出embeddings中第1,3,5行,组成一个tensor返回。
embedding_lookup不是简单的查表,id对应的向量是可以训练的,训练参数个数应该是 category num*embedding size,也就是说lookup是一种全连接层。
看一段代码:
#!/usr/bin/env/python
# coding=utf-8
import tensorflow as tf
import numpy as np
# 定义一个未知变量input_ids用于存储索引
input_ids = tf.placeholder(dtype=tf.int32, shape=[None])
# 定义一个已知变量embedding,是一个5*5的对角矩阵
# embedding = tf.Variable(np.identity(5, dtype=np.int32))
# 或者随机一个矩阵
embedding = a = np.asarray([[0.1, 0.2, 0.3], [1.1, 1.2, 1.3], [2.1, 2.2, 2.3], [3.1, 3.2, 3.3], [4.1, 4.2, 4.3]])
# 根据input_ids中的id,查找embedding中对应的元素
input_embedding = tf.nn.embedding_lookup(embedding, input_ids)
sess = tf.InteractiveSession()
sess.run(tf.global_variables_initializer())
# print(embedding.eval())
print(sess.run(input_embedding, feed_dict={input_ids: [1, 2, 3, 0, 3, 2, 1]}))
tf.contrib.layers.xavier_initializer
https://blog.csdn.net/qq_30868235/article/details/80939462
xavier_initializer(
uniform=True,
seed=None,
dtype=tf.float32
)
该函数返回一个用于初始化权重的初始化程序 “Xavier” 。
这个初始化器是用来保持每一层的梯度大小都差不多相同。
参数:
uniform: 使用uniform或者normal分布来随机初始化。
seed: 可以认为是用来生成随机数的seed
dtype: 只支持浮点数。
返回值:
初始化权重矩阵
tf.reduce_mean()
1.1 函数声明
tf.reduce_mean(input_tensor, reduction_indices=None, keep_dims=False, name=None)
- 1
1.2 函数说明
Computes the mean of elements across dimensions of a tensor.
Reduces input_tensor along the dimensions given in reduction_indices. Unless keep_dims is true, the rank of the tensor is reduced by 1 for each entry in reduction_indices. If keep_dims is true, the reduced dimensions are retained with length 1.
If reduction_indices has no entries, all dimensions are reduced, and a tensor with a single element is returned.
1.3 参数说明
Args:
input_tensor: The tensor to reduce. Should have numeric type.
reduction_indices: The dimensions to reduce. If None (the default), reduces all dimensions.
keep_dims: If true, retains reduced dimensions with length 1.
name: A name for the operation (optional).
1.4 返回值
Returns:
The reduced tensor.
1.5 举例
For example:
# 'x' is [[1., 1.]
# [2., 2.]]
tf.reduce_mean(x) ==> 1.5
tf.reduce_mean(x, 0) ==> [1.5, 1.5]
tf.reduce_mean(x, 1) ==> [1., 2.]
tf.reduce_sum
-
reduce_sum(
-
input_tensor,
-
axis=None,
-
keep_dims=False,
-
name=None,
-
reduction_indices=None
-
)
input_tensor:表示输入
axis:表示在那个维度进行sum操作。
keep_dims:表示是否保留原始数据的维度,False相当于执行完后原始数据就会少一个维度。
reduction_indices:为了跟旧版本的兼容,现在已经不使用了。
官方的例子:
-
# 'x' is [[1, 1, 1]
-
# [1, 1, 1]]
-
tf.reduce_sum(x) ==> 6
-
tf.reduce_sum(x, 0) ==> [2, 2, 2]
-
tf.reduce_sum(x, 1) ==> [3, 3]
-
tf.reduce_sum(x, 1, keep_dims=True) ==> [[3], [3]]
-
tf.reduce_sum(x, [0, 1]) ==> 6
tf.transpose() 转置
这里的Perm是指定的转置维度,不指定,默认二维的行列交换
transpose(
a,
perm=None,
name='transpose'
)
For example:默认情况
x = tf.constant([[1, 2, 3], [4, 5, 6]])
tf.transpose(x) # [[1, 4]
# [2, 5]
# [3, 6]]
- 这里,指定1维与0维交换,也就是与默认情况相同
tf.transpose(x, perm=[1, 0]) # [[1, 4]
# [2, 5]
# [3, 6]]
- perm在多维情况下更有用,如下
# 'perm' is more useful for n-dimensional tensors, for n > 2
x = tf.constant([[[ 1, 2, 3],
[ 4, 5, 6]],
[[ 7, 8, 9],
[10, 11, 12]]])
# Take the transpose of the matrices in dimension-0
tf.transpose(x, perm=[0, 2, 1]) # [[[1, 4],
# [2, 5],
# [3, 6]],
# [[7, 10],
# [8, 11],
# [9, 12]]]
- 本来矩阵是[2,2,3]的,但perm指定第二维和第三维交换,转后变为了[2,3,2]
- a的转置是根据 perm 的设定值来进行的。
返回数组的 dimension(尺寸、维度) i与输入的 perm[i]的维度相一致。如果未给定perm,默认设置为 (n-1…0),这里的 n 值是输入变量的 rank 。
tf.reshape()
reshape(tensor, shape, name=None)
主要通过改变张量形状,可以从高维变低维,也可以从低维变高维;
a = tf.Variable(initial_value=[[1,2,3],[4,5,6]]) ==> shape:[2,3]
b = tf.Variable(initial_value=[[[1,2,3],[4,5,6]],[[7,8,9],[1,0,2]]]) ==> shape:[2,2,3]
a_1 = tf.reshape(a,[2,1,1,3]) ==> [[[[1,2,3]]],[[[4,5,6]]]]
a_2 = tf.reshape(a,[2,1,3]) ==> [[[1,2,3]],[[4,5,6]]]
b_1 = tf.reshape(b,[2,2,1,3]) ==> [[[[1,2,3]],[[4,5,6]]],[[[7,8,9]],[[1,0,2]]]]
new_1 = tf.concat([b_1,a_1],1)
new_2 = tf.reshape(tf.concat([b,a_2],1),[2,3,1,3])
"""
new_1:
[[[[1 2 3]]
[[4 5 6]]
[[1 2 3]]]
[[[7 8 9]]
[[1 0 2]]
[[4 5 6]]]]
new_2;
[[[[1 2 3]]
[[4 5 6]]
[[1 2 3]]]
[[[7 8 9]]
[[1 0 2]]
[[4 5 6]]]]
tf.train.exponential_decay
用一个例子说明,这个是用来使学习率自动下降的东西
from ctypes import *
import tensorflow as tf
import matplotlib.pyplot as plt
alpha = 0.1
decay_rate = 0.96
decay_step = 1000
global_ = tf.Variable(tf.constant(0))
global_step = 10000
learn_rate = tf.train.exponential_decay(alpha,global_,decay_step,decay_rate,staircase=True)
c = []
with tf.Session() as sess:
for i in range(global_step):
tc = sess.run(learn_rate,feed_dict={global_:i})
c.append(tc)
plt.figure(1)
plt.plot(range(global_step),c,'r-')
plt.show()