web-dev-qa-db-ja.com

テンソルフローで転置を並べ替える方法は?

docs から:

aを転置します。パーマに従って次元を並べ替えます。

返されるテンソルの次元iは、入力次元perm[i]に対応します。 permが指定されていない場合、(n-1 ... 0)に設定されます。ここで、nは入力テンソルのランクです。したがって、デフォルトでは、この操作は2次元入力テンソルに対して通常の行列転置を実行します。

しかし、入力テンソルをどのようにスライスする必要があるかは、まだはっきりしていません。例えば。ドキュメントからも:

tf.transpose(x, perm=[0, 2, 1]) ==> [[[1  4]
                                      [2  5]
                                      [3  6]]

                                     [[7 10]
                                      [8 11]
                                      [9 12]]]

perm=[0,2,1]が1x3x2テンソルを生成するのはなぜですか?

試行錯誤の後:

twothreefour = np.array([ [[1,2,3,4], [5,6,7,8], [9,10,11,12]] , 
                        [[13,14,15,16], [17,18,19,20], [21,22,23,24]] ])
twothreefour

[アウト]:

array([[[ 1,  2,  3,  4],
        [ 5,  6,  7,  8],
        [ 9, 10, 11, 12]],

       [[13, 14, 15, 16],
        [17, 18, 19, 20],
        [21, 22, 23, 24]]])

そして、私がそれを転置すると:

fourthreetwo = tf.transpose(twothreefour) 
with tf.Session() as sess:
    init = tf.initialize_all_variables()
    sess.run(init)
    print (fourthreetwo.eval())

私は4x3x2から2x3x4を取得し、それは論理的に聞こえます。

[アウト]:

[[[ 1 13]
  [ 5 17]
  [ 9 21]]

 [[ 2 14]
  [ 6 18]
  [10 22]]

 [[ 3 15]
  [ 7 19]
  [11 23]]

 [[ 4 16]
  [ 8 20]
  [12 24]]]

しかし、permパラメーターの出力を使用すると、実際に何が得られているのかわかりません。

twofourthree = tf.transpose(twothreefour, perm=[0,2,1]) 
with tf.Session() as sess:
    init = tf.initialize_all_variables()
    sess.run(init)
    print (threetwofour.eval())

[アウト]:

[[[ 1  5  9]
  [ 2  6 10]
  [ 3  7 11]
  [ 4  8 12]]

 [[13 17 21]
  [14 18 22]
  [15 19 23]
  [16 20 24]]]

perm=[0,2,1]が2x3x4から2x4x3行列を返す理由

perm=[1,0,2]でもう一度お試しください:

threetwofour = tf.transpose(twothreefour, perm=[1,0,2]) 
with tf.Session() as sess:
    init = tf.initialize_all_variables()
    sess.run(init)
    print (threetwofour.eval())

[アウト]:

[[[ 1  2  3  4]
  [13 14 15 16]]

 [[ 5  6  7  8]
  [17 18 19 20]]

 [[ 9 10 11 12]
  [21 22 23 24]]]

perm=[1,0,2]が2x3x4から3x2x4を返す理由

permパラメータがnp.shapeを取り、配列の形状に基づく要素に基づいてテンソルを転置していることを意味しますか?

つまり:

_size = (2, 4, 3, 5)
randarray = np.random.randint(5, size=_size)

shape_idx = {i:_s for i, _s in enumerate(_size)}

randarray_t_func = tf.transpose(randarray, perm=[3,0,2,1]) 
with tf.Session() as sess:
    init = tf.initialize_all_variables()
    sess.run(init)
    tranposed_array = randarray_t_func.eval()
    print (tranposed_array.shape)

print (Tuple(shape_idx[_s] for _s in [3,0,2,1]))

[アウト]:

(5, 2, 3, 4)
(5, 2, 3, 4)
17
alvas

permは次元を入れ替えていると思います。たとえば、perm=[0,2,1]dim_0 -> dim_0, dim_1 -> dim_2, dim_2 -> dim_1の略です。したがって、2Dテンソルの場合、perm=[1,0]は単なる行列転置です。これはあなたの質問に答えますか?

30
maxymoo
A=[2,3,4] matrix, using perm(1,0,2) will get B=[3,2,4].

説明:

Index=(0,1,2)
A    =[2,3,4]
Perm =(1,0,2)
B    =(3,2,4)  --> Perm 1 from Index 1 (3), Perm 0 from Index 0 (2), Perm 2 from Index 2 (4) --> so get (3,2,4)
3
Kew Hsein