TensowFlow

Question 1

Writing and running programs in TensorFlow has the following steps:

Answer 1

Create Tensors (variables) that are not yet executed/evaluated.
Write operations between those Tensors.
Initialize your Tensors.
Create a Session.
Run the Session. This will run the operations you’d written above

Question 2

What is a placeholder?

Answer 2

A placeholder is an object whose value you can specify only later.

Question 3

How do you put values in a placholder?

Answer 3

You use the function feed_dict, feed_dict = {x: 3}

Question 4

How do you do a matrix multiplication in tensforflow?

Answer 4

tf.matmul(…, …) to do a matrix multiplication

Question 5

Addition in tensorflow?

Answer 5

tf.add(…, …) to do an addition

Question 6

random initialization in python

Answer 6

np.random.randn(…) to initialize randomly

Question 7

how would you define a constant X that has shape (3,1)?

Answer 7

X = tf.constant(np.random.randn(3,1), name = “X”)

Question 8

How would you declare a placeholder?

Answer 8

tf.placeholder(
    dtype,
    shape=None,
    name=None
)
E.g. x = tf.placeholder(tf.float32, shape=(1024, 1024))

Question 9

How would you initialize parameters in tensorflow?

Answer 9

tf.get_variable(
    name,
    shape=None,
    dtype=None,
    initializer=None,
    regularizer=None,
    trainable=True,
    collections=None,
    caching_device=None,
    partitioner=None,
    validate_shape=True,
    use_resource=None,
    custom_getter=None,
    constraint=None
)

Question 10

How to do a 2d convolution in TF?

Answer 10

tf.nn.conv2d(X,W1, strides = [1,s,s,1], padding = ‘SAME’)

Question 11

how to maxpool in TF?

Answer 11

tf.nn.max_pool(A, ksize = [1,f,f,1], strides = [1,s,s,1], padding = ‘SAME’):

Question 12

how to ReLU in TF?

Answer 12

tf.nn.relu(Z1

Question 13

How to flatten a layer in TF?

Answer 13

tf.contrib.layers.flatten(P)

Question 14

How to do a fully connected layer in TF?

Answer 14

tf.contrib.layers.fully_connected(F, num_outputs)

Question 15

How do you softmax with several labels in TF?

Answer 15

tf.nn.softmax_cross_entropy_with_logits(logits = Z3, labels = Y)

Question 16

How do you setup the opimizer in TF?

Answer 16

optimizer = tf.train.GradientDescentOptimizer(learning_rate = learning_rate).minimize(cost)

Question 17

How do you run the session (on optimizer and cost) in TF?

Answer 17

_ , c = sess.run([optimizer, cost], feed_dict={X: minibatch_X, Y: minibatch_Y})