Tukukkk

I am trying to write a program in python using tensorflow to fit data with different patterns using a neural network with a hidden layer. I am facing an error with the code which states that the variable b2 is uninitialized. But I have initialized it already and dont understand what I am missing here.

This is part of an assignment (the dataset can be understood here) and got stuck here while solving it.

The link to the colab notebook is here.

The initialization lines look like

W1 = tf.Variable(np.random.uniform(low=-0.01, high=0.01, size=(hidden, 2)), name="W1")

b1 = tf.Variable(np.random.uniform(low=-0.01, high=0.01, size=(hidden, 1)), name="b1")

W2 = tf.Variable(np.random.uniform(low=-0.01, high=0.01, size=(classes, hidden)), name="W2")

b2 = tf.Variable(np.random.uniform(low=-0.01, high=0.01, size=(classes, 1)), name="b2")

The code snippet that shows the computation graph is below.

operation = "ReLU" # "Sigmoid"

o = tf.add(tf.matmul(W1, p), b1)

# ReLU or Sigmoid

if operation == "ReLU":

  z = tf.zeros((hidden, 1), dtype=tf.float64)

  output = tf.maximum(o, z)

else:

  output = tf.sigmoid(o)

foutput = tf.add(tf.matmul(W2, output), b2)

crossentropy = tf.log(tf.exp(foutput) / tf.reduce_sum(tf.exp(foutput), 0))

init = tf.initialize_all_variables()

with tf.Session() as sess:

  ce = sess.run([crossentropy], feed_dict={p : inputs, t : targets, lr : 0.01})

print(ce)

Error message

FailedPreconditionError                   Traceback (most recent call last)

/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py in _do_call(self, fn, *args)

   1333     try:

-> 1334       return fn(*args)

   1335     except errors.OpError as e:



/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py in _run_fn(feed_dict, fetch_list, target_list, options, run_metadata)

   1318       return self._call_tf_sessionrun(

-> 1319           options, feed_dict, fetch_list, target_list, run_metadata)

   1320 



/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py in _call_tf_sessionrun(self, options, feed_dict, fetch_list, target_list, run_metadata)

   1406         self._session, options, feed_dict, fetch_list, target_list,

-> 1407         run_metadata)

   1408 



FailedPreconditionError: Attempting to use uninitialized value b2

     [[{{node b2/read}} = Identity[T=DT_DOUBLE, _device="/job:localhost/replica:0/task:0/device:CPU:0"](b2)]]

You also need to run the init function in the graph before computing cross entropy:

with tf.Session() as sess:

  sess.run(init)

  ce = sess.run([crossentropy], feed_dict={p : inputs, t : targets, lr : 0.01})

Also, tf.initialize_all_variables() is deprecated. Use tf.global_variables_initializer() instead.