Python/numpy/tensorflow, the matrix row and column operations, subscript is how to go?

The Ndarrray in List/tuple,numpy in Python and the tensor in TensorFlow.

In Python, List/tuple understands that a sequence of data is understood only from the memory point of view, not the number of mathematical bids, vectors, and tensor.

From the Python memory point of view, is a numeric value, length 1, and is not a sequence;

From the perspective of NumPy and TensorFlow mathematics, it is a scalar, shape is (), its axis is 0;

[1,2,3,4,5,6]

From the Python memory perspective, it is a sequence of 1*6 or a length of 6;

From NumPy and TensorFlow mathematical perspective, is a one-dimensional vector, and is a column vector, Shape is (6,), its axis is 1; note: one-dimensional vector, column vector, its axis is 1, the axis subscript 0, representing the first axis;

[[A] [[4,5,6]]

From the Python memory point of view, is a two-dimensional array 2*3,2 column 3 rows; Note: Whether Python or Numpy/tensorflow are listed before, that is, a Shang is considered a column vector;

From the Numpy/tensorflow mathematical point of view, is a two-dimensional tensor, shape is (2,3), its axis is 2; note: two-dimensional vector, the first axis is a column, subscript is 0, the second axis is a row, subscript is 1;

Example:

Import OS, sys

Import NumPy

Import TensorFlow as TF

A = [[1,2,3],[4,5,6]]

b = tf. Variable (A, Dtype=tf.float32)

init = Tf.global_variables_initializer ()

Sess = tf. Session ()

Sess.run (INIT)

D1 = Tf.reduce_mean (b)

D2 = Tf.reduce_mean (b, 0)

D3 = Tf.reduce_mean (b, 1)

Sess.run ([B, B[0,:], b[:, 0]])

# [Array ([[1., 2., 3.],

# [4., 5., 6.]], Dtype=float32),

# array ([1., 2., 3.], Dtype=float32),

# array ([1., 4.], Dtype=float32)]

Sess.run ([D1, D2, D3])

# [3.5,

# array ([2.5, 3.5, 4.5], dtype=float32),

# Array ([2., 5.], Dtype=float32)]

Note:

[B, B[i,:], b[:, I]]

B[i,:]: denotes the first axis (column) of the B matrix, and the column I corresponds to all elements;

B[:,i]: Represents the second axis of the B matrix (row), the line I corresponds to all the elements;

D1=tf.reduce_mean (b): means the mean value of all elements of the matrix;

D2=tf.reduce_mean (b, 0): represents the projection on the first axis (column) direction of the Matrix, that is, the mean value on each row;

D3=tf.reduce_mean (b, 1): represents the projection on the second axis (row) direction of the Matrix, that is, the mean value of each column;

View this URL: http://www.cnblogs.com/silence-tommy/p/6554986.html

There are questions on this website:

Tf.reduce_mean (x) ==> 2.5 #如果不指定第二个参数, then the average of all the elements

Tf.reduce_mean (x, 0) ==> [2., 3.] #指定第二个参数为0, the elements of the first dimension are averaged, that is, each column averages

Tf.reduce_mean (x, 1) ==> [1.5, 3.5] #指定第二个参数为1, the elements of the second dimension are averaged, that is, each row is averaged

These two underlined callouts are wrong. Clearly no understanding of the concept of matrix axes;

(Sorry, this site blogger, just see, so only from the perspective of development analysis, not intentionally, please understand!) )

Python/numpy/tensorflow, the matrix row and column operations, subscript is how to go?