Tensorflow While loop with Variable Creation

Question

Tensorflow While loop with Variable Creation Code here :

x = tf.Variable(100)    
c = tf.Constant(2)    
n = 100    
loops = 50

l1 = tf.Variable(np.random.random(n))    
c1 = tf.Variable(np.random.random(n))     
x = tf.multiply(c1,tf.exp(-(x-l1)/c))

l2 = tf.Variable(np.random.random(n))
c2 = tf.Variable(np.random.random(n)) 
x = tf.multiply(c2,tf.exp(-(x-l2)/c))

l3 = tf.Variable(np.random.random(n))
c3 = tf.Variable(np.random.random(n))
x = tf.multiply(c3,tf.exp(-(x-l3)/c))

.....

.....

l50 = tf.Variable(np.random.random(n))
c50 = tf.Variable(np.random.random(n))
x = tf.multiply(c50,tf.exp(-(x-l50)/c))

So, I want to do that in while loop in tensorflow as:

while loop(i from 1 to 50):
   l[i] = tf.Variable(np.random.random(n))
   c[i] = tf.Variable(np.random.random(n))
   x = tf.multiply(c[i],tf.exp(-(x-l[i])/c))

How can I achieve this in tensorflow. Thanking you

You can format blocks of code by indenting it all by four spaces, or by highlighting it all and pressing the {} (format code) button. You don't need to individually backtick every line. — khelwood
– khelwood, Commented Jul 16, 2018 at 10:31
Can you tell the actual formula, might help to come up with a vectorized implementation. — Mahesh
– Mahesh, Commented Jul 16, 2018 at 12:14

Mohan Radhakrishnan · Accepted Answer · 2018-07-18 09:35:53Z

1

Can you feed it like this ?

l = tf.placeholder(tf.float32, shape=[None, ])
c = tf.placeholder(tf.float32, shape=[None, ])

sess = tf.Session()
sess.run(tf.global_variables_initializer())

x = tf.multiply(l,c)   #Assume a formula

for i in range(50) :
    arr = np.random.random_sample((i,))
    print (arr)
    sess.run(x,feed_dict={l :arr,c :arr})

Based on your comment I've only attempted. This isn't the exact answer.

results = tf.TensorArray(dtype=tf.float32, size=100)

lvalues = tf.TensorArray(dtype=tf.float32, size=50)
cvalues = tf.TensorArray(dtype=tf.float32, size=50)

sess = tf.Session()
sess.run(tf.global_variables_initializer())

def my_func(x):
    a = np.random.random_sample((x,))
    return a

input = tf.placeholder(tf.int32)
y = tf.py_func(my_func, [input], tf.float64)
for i in range(50) :
    r = sess.run(y,feed_dict={input : i})
    lvalues.write(i,y)
    cvalues.write(i,y)
    print(y)
    print(r)

edited Jul 18, 2018 at 9:35

answered Jul 16, 2018 at 12:26

Mohan Radhakrishnan

3,2385 gold badges35 silver badges46 bronze badges

Sign up to request clarification or add additional context in comments.

2 Comments

pku Over a year ago

No SIr, Actually, I want to use computation graph for these variable.

Mohan Radhakrishnan Over a year ago

Please update your question with some more details so that others can attempt.

Collectives™ on Stack Overflow

Tensorflow While loop with Variable Creation

1 Answer 1

2 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

2 Comments

Your Answer

Sign up or log in

Post as a guest

Related