We have learned how to create a random orthogonal matrix by scipy in python. In this tutorial, we will introduce how to creat a random orthogonal matrix using tensorflow.
To create a random orthogonal matrix using scipy. we can read:
Python Create a Random Orthogonal Matrix: A Beginner Guide – Python Tutorial
How to create a random orthogonal matrix using tensorflow?
We can use tf.orthogonal_initializer() to implement it.
Here is an example:
import tensorflow as tf import numpy as np w_ = tf.Variable(tf.orthogonal_initializer()([5, 10]), name = 'weight_')
where w_ is a random orthogonal matrix.
In order to evaluate whether w_ is an orthogonal matrix or not, we can do as follows:
s_ = tf.matmul(w_, w_, transpose_b = True)
Then we can output w_ and s_
with tf.Session() as sess:
sess.run(tf.global_variables_initializer())
sess.run(tf.local_variables_initializer())
np.set_printoptions(precision=3, suppress=True)
print(sess.run(w_))
print(sess.run(s_))
Run this code, we can find w_ and s_ are:
[[ 0.345 0.044 -0.055 0.208 0.262 0.31 -0.809 0.034 -0.087 0.072] [-0.043 0.269 0.627 -0.297 -0.201 -0.414 -0.348 -0.106 -0.219 -0.228] [ 0.295 -0.186 0.029 0.105 -0.51 -0.072 0.039 -0.18 -0.389 0.645] [ 0.232 -0.295 -0.391 -0.368 0.374 -0.365 0.032 -0.07 -0.507 -0.186] [ 0.419 -0.001 0.155 -0.127 0.012 0.365 0.212 -0.721 0.122 -0.268]] [[ 1. -0. 0. 0. 0.] [-0. 1. 0. 0. -0.] [ 0. 0. 1. 0. 0.] [ 0. 0. 0. 1. -0.] [ 0. -0. 0. -0. 1.]]
It means that we have created a random orthogonal matrix w_ using tensorflow.