How to duplicate operations & placeholders in Tensorflow -

suppose have 2 neural network model defined each 1 input placeholder , 1 output tensor. these 2 outputs need 3 separate values.

inputs: i1, i2, outputs: o1, o2 = 1 b = 2  v1 = session.run(o1, feed_dict={i1: a}) v2 = session.run(o1, feed_dict={i1: b}) v3 = session.run(o2, feed_dict={i2: a}) 

the problem need feed these 3 values loss function can't above. need do

loss = session.run(l, feed_dict={i1: a, i1: b, i2:a }) 

i don't think can if still have ambiguity in later operations since o1 input i1 used differently o1 input i2.

i think solved having 2 input placeholders , 2 outputs in first neural network. given have model there way restructure inputs , outputs can accommodate this?

visually want turn

i1 ---- (model) ----- o1  

into

i1a                          o1a   \                         /    \                       /     x ----- (model) ----- x            /                       \   /                         \ i1b                          o1b 

your intuition right, have create 2 different placeholders i1a , i1b network 1, 2 outputs o1a , o1b. visuals great here proposition:

i1a  ----- (model) -----  o1a               |                     shared weights                                                 |             i1b  ----- (model) -----  o1b 

---

the proper way duplicate network using tf.get_variable() every variable reuse=true.

def create_variables():   tf.variable_scope('model'):     w1 = tf.get_variable('w1', [1, 2])     b1 = tf.get_variable('b1', [2])  def inference(input):   tf.variable_scope('model', reuse=true):     w1 = tf.get_variable('w1')     b1 = tf.get_variable('b1')     output = tf.matmul(input, w1) + b1   return output  create_variables()  i1a = tf.placeholder(tf.float32, [3, 1]) o1a = inference(i1a)  i1b = tf.placeholder(tf.float32, [3, 1]) o1b = inference(i1b)  loss = tf.reduce_mean(o1a - o1b)   tf.session() sess:   sess.run(tf.initialize_all_variables())   sess.run(loss, feed_dict={i1a: [[0.], [1.], [2.]], i1b: [[0.5], [1.5], [2.5]]})