矩阵张量流中的链接变量

Question

我想知道在TensorFlow中是否有一种简单的方法来约束矩阵中的变量。

作为一个玩具示例，我写了一段代码，希望我的input_matrix能够汇聚到[[2., 1.], [1., 2.]]。

import tensorflow as tf

sess = tf.Session()
v1 = tf.Variable(1.)
v2 = tf.Variable(2.)
#Here I specify that some of the variables in the matrix must have the same values, but it obviously doesn't work since TensorFlow variables need to be initialized before being used
input_matrix = tf.Variable([[v1, v2], [v2, v1]])
objective_matrix = tf.constant([[0., 1.], [1., 4.]])
optimizer = tf.train.GradientDescentOptimizer(1e-1)
cost = tf.reduce_sum(tf.square(tf.subtract(objective_matrix, input_matrix)))
train_step = optimizer.minimize(cost)
sess.run(tf.global_variables_initializer())
for _ in range(100):
    sess.run(train_step)

那么，是否可以强制矩阵的某些元素相等或至少是渐变？