tensorflow学习笔记（二）：tensor 变换

#对于2-D #所有的reduce_...，如果不加axis的话，都是对整个矩阵进行运算 tf.reduce_sum(a, 1） #对axis1 tf.reduce_mean(a,0) #每列均值 

第二个参数是axis，如果为0的话，

r

e

s

[

i

]

=

∑

j

a

[

j

,

i

]

res[i] = \sum_{j} a[j,i]

res[i]=∑ja[j,i]即（

r

e

s

[

i

]

=

∑

a

[

:

,

i

]

res[i] = \sum a[:,i]

res[i]=∑a[:,i]）， 如果是1的话，

r

e

s

[

i

]

=

∑

j

a

[

i

,

j

]

res[i] = \sum_{j} a[i,j]

res[i]=∑ja[i,j]
NOTE:返回的都是行向量,（axis等于几，就是对那维操作,i.e.:沿着那维操作, 其它维度保留）

#关于concat，可以用来进行降维 3D->2D , 2D->1D tf.concat(concat_dim, data) #arr = np.zeros([2,3,4,5,6]) In [6]: arr2.shape Out[6]: (2, 3, 4, 5) In [7]: np.concatenate(arr2, 0).shape Out[7]: (6, 4, 5)   :(2*3, 4, 5) In [9]: np.concatenate(arr2, 1).shape Out[9]: (3, 8, 5)   :(3, 2*4, 5) #tf.concat() t1 = [[1, 2, 3], [4, 5, 6]] t2 = [[7, 8, 9], [10, 11, 12]] # 将t1, t2进行concat，axis为0，等价于将shape=[2, 2, 3]的Tensor concat成 #shape=[4, 3]的tensor。在新生成的Tensor中tensor[:2,:]代表之前的t1 #tensor[2:,:]是之前的t2 tf.concat(0, [t1, t2]) ==> [[1, 2, 3], [4, 5, 6], [7, 8, 9], [10, 11, 12]]  # 将t1, t2进行concat，axis为1，等价于将shape=[2, 2, 3]的Tensor concat成 #shape=[2, 6]的tensor。在新生成的Tensor中tensor[:,:3]代表之前的t1 #tensor[:,3:]是之前的t2 tf.concat(1, [t1, t2]) ==> [[1, 2, 3, 7, 8, 9], [4, 5, 6, 10, 11, 12]] 

concat是将list中的向量给连接起来，axis表示将那维的数据连接起来，而其他维的结构保持不变

#squeeze 降维 维度为1的降掉 tf.squeeze(arr, []) 降维， 将维度为1 的降掉 arr = tf.Variable(tf.truncated_normal([3,4,1,6,1], stddev=0.1)) arr2 = tf.squeeze(arr, [2,4]) arr3 = tf.squeeze(arr) #降掉所以是1的维  #split tf.split(split_dim, num_split, value, name='split') # 'value' is a tensor with shape [5, 30] # Split 'value' into 3 tensors along dimension 1 split0, split1, split2 = tf.split(1, 3, value) tf.shape(split0) ==> [5, 10]  #embedding mat = np.array([1,2,3,4,5,6,7,8,9]).reshape((3,-1)) ids = [[1,2], [0,1]] res = tf.nn.embedding_lookup(mat, ids) res.eval() array([[[4, 5, 6],         [7, 8, 9]],         [[1, 2, 3],         [4, 5, 6]]])  #扩展维度，如果想用广播特性的话，经常会用到这个函数 # 't' is a tensor of shape [2] #一次扩展一维 shape(tf.expand_dims(t, 0)) ==> [1, 2] shape(tf.expand_dims(t, 1)) ==> [2, 1] shape(tf.expand_dims(t, -1)) ==> [2, 1] # 't2' is a tensor of shape [2, 3, 5] shape(tf.expand_dims(t2, 0)) ==> [1, 2, 3, 5] shape(tf.expand_dims(t2, 2)) ==> [2, 3, 1, 5] shape(tf.expand_dims(t2, 3)) ==> [2, 3, 5, 1]  

tf.slice()

tf.slice(input_, begin, size, name=None) 

先看例子

import tensorflow as tf import numpy as np sess = tf.InteractiveSession() a = np.array([[1,2,3,4,5],[4,5,6,7,8],[9,10,11,12,13]]) tf.slice(a,[1,2],[-1,2]).eval()  #array([[ 6,  7], #       [11, 12]]) 

理解tf.slice()最好是从返回值上去理解，现在假设input的shape是[a1, a2, a3], begin的值是[b1, b2, b3],size的值是[s1, s2, s3],那么tf.slice()返回的值就是 input[b1:b1+s1, b2:b2+s2, b3:b3+s3]。
如果

s

i

=

−

1

s_i=-1

si=−1 ，那么 返回值就是 input[b1:b1+s1,..., bi: ,...]

注意：input[1:2] 取不到input[2]

tf.stack()

tf.stack(values, axis=0, name='stack') 

将 a list of R 维的Tensor堆成 R+1维的Tensor。
Given a list of length N of tensors of shape (A, B, C);
if axis == 0 then the output tensor will have the shape (N, A, B, C)

这时 res[i,:,:,:] 就是原 list中的第 i 个 tensor 

. if axis == 1 then the output tensor will have the shape (A, N, B, C).

这时 res[:,i,:,:] 就是原list中的第 i 个 tensor 

Etc.

# 'x' is [1, 4] # 'y' is [2, 5] # 'z' is [3, 6] stack([x, y, z]) => [[1, 4], [2, 5], [3, 6]]  # Pack along first dim. stack([x, y, z], axis=1) => [[1, 2, 3], [4, 5, 6]] 

tf.pad

tf.pad(tensor, paddings, mode="CONSTANT", name=None) 

• tensor: 任意shape的tensor，维度 Dn
• paddings: [Dn, 2] 的 Tensor, Padding后tensor的某维上的长度变为padding[D,0]+tensor.dim_size(D)+padding[D,1]
• mode: CONSTANT表示填0, REFLECT表示反射填充,SYMMETRIC表示对称填充。

tf.gather()

tf.gather(params, indices, validate_indices=None, name=None) 

indices must be an integer tensor of any dimension (usually 0-D or 1-D). Produces an output tensor with shape indices.shape + params.shape[1:]

# Scalar indices, 会降维 output[:, ..., :] = params[indices, :, ... :]  # Vector indices output[i, :, ..., :] = params[indices[i], :, ... :]  # Higher rank indices，会升维 output[i, ..., j, :, ... :] = params[indices[i, ..., j], :, ..., :] """ indices = [2, 0, 3] output[0] <-- param[2]  output[1] <-- param[0] output[2] <-- param[3]  indices = [[1, 2], [2, 3]] output[0, 0] <-- param[1] output[0, 1] <-- param[2] output[1, 0] <-- param[2] output[1, 1] <-- param[3] """ 

tf.gather_nd

tf.gather_nd(     params, indices, name=None, batch_dims=0 ) indices = [[1, 2], [2, 3]] output[0] <-- param[1, 2] output[1] <-- param[2, 3] . # 注意对比与 tf.gather之间的区别  indices = [[[1, 2]],  			[[2, 3]]] output[0, 0] <-- param[1, 2] output[1, 0] <-- param[2, 3] . # 注意对比与 tf.gather之间的区别 

tf.scatter_nd

scatter_nd(     indices,     updates,     shape,     name=None ) """ shape: 最终结果的shape updates: 数据源 indices: 数据源相应未知的数据放到结果的什么位置 """ # 文档中废话那么多，可以总结成两个式子 # res[*indice[i,j,..,z], ...] = updates[i,j,..,z,...] # len([i,j,..,z]) = indice.rank-1.    [i,j,...,z]表示原始未知。*indice[i,j,..,z] 表示目标位置！ # 在这里 rank 表示 几维， a=[1,2,3], a.rank=1,  b = [[1,2], [2,3], [3,4]], b.rank=2 """ indices = [ [1, 2, 3], 			[2, 3, 4]	 		  ] 意味着  	updates[0] --> res[1, 2, 3]  	updates[1] --> res[2, 3, 4]  indices = [ [[1, 2, 3], [7, 8, 9]], 			[[2, 3, 4], [9, 10, 11]] 		] 意味着: 	updates[0, 0] --> res[1, 2, 3] 	updates[0, 1] --> res[7, 8, 9] 	updates[1, 0] --> res[2, 3, 4] 	updates[1, 1] --> res[9, 10, 11] """