Basics

Question 1

How do you create a session?

Answer 1

sess = tf.Session()

Question 2

What do you need to do to use TensorFlow in Python?

Answer 2

import tensorflow as tf

Question 3

How do you create a constant in TensorFlow?

Answer 3

a = tf.constant(3.0, dtype=tf.float32)

Question 4

How do you run a TensorFlow session?

Answer 4

sess.run([node1, node2])

Question 5

How can you print a session to the prompt?

Answer 5

print(sess.run([node1, node2])

Question 6

How do you add values in TensorFlow?

Answer 6

c = a + b
c = tf.add(a, b)

Question 7

How do you add inputs to TensorFlow?

Answer 7

a = tf.placeholder(tf.float32)
b = ft.placeholder(tf.float32)
c = a + b
sess.run(c, {a: 3, b: 4})
will give 7
to print to the console use
print(sess.run(c, {a: 3, b: 4}))

Question 8

How do you multiply a and b in TensorFlow

Answer 8

c = a * b

Question 9

How can you stack node in TensorFlow?

Answer 9

c = a + b
d = c * 3

Question 10

How do you create a variable in TensorFlow?

Answer 10

W = tf.Variable([0.3], dtype=tf.float32)

Question 11

Create a linear model with TensorFlow

Answer 11

W = tf.Variable([0.3], dtype=tf.float32)
b = tf.Variable([-0.3], dtype=tf.float32)
x = tf.placeholder(tf.float32)
linear_model = W * x + b

Question 12

How do you initialize variables in TensorFlow?

Answer 12

init = tf.global_variables_initializer()

sess.run(init)

Question 13

How can you add the contents of a vector V together in TensorFlow?

Answer 13

total = tf.reduce_sum(V)

Question 14

Write a MSD error function in TensorFlow.

Answer 14

y = tf.placeholder(tf.float32)
squared_deltas = tf.square(linear_model - y)
loss = tf.reduce_sum(squared_deltas)

to run the model and print the output use something like
print( sess.run( loss, { x: [1, 2, 3, 4], y: [0, -1, -2, -3] } ) )

Question 15

How can you set the value of a variable in TensorFlow?

Answer 15

fixW = tf.assign(W, [-1.0])

sess.run(fixW)

Question 16

How do you setup a gradient descent optimizer in TensorFlow?

Answer 16

optim = tf.train.GradientDescentOptimizer(0.01)
train = optim.minimize(loss)

Question 17

How do you run the TensorFlow optimizer once its setup?

Answer 17

sess.run(init) # reset values to defaults.
for i in range(1000):
sess.run(train, {x: [1, 2, 3, 4], y: [0, -1, -2, -3]})

print(sess.run([W, b]))

Question 18

How do you check if a file is executable?

Answer 18

ls -la

Question 19

How can you make a file executable?

Answer 19

chmod +x fname

Question 20

How can you produce a series of equally spaced numbers with numpy?

Answer 20

x = np.linspace(-3.0, 3.0, 100)

Question 21

How can you produce a series of equally spaced numbers with TensorFlow?

Answer 21

x = tf.linspace(-3.0, 3.0, 100)

Question 22

How do you start an interactive session in TensorFlow?

Answer 22

sess = tf.InteractiveSession()

Question 23

How do you get the default graph in TensorFlow?

Answer 23

g = tf.get_default_graph()

Question 24

How can you get a list of all the operations that have been added to a graph g?

Answer 24

[op.name for op in g.get_operations()]

Question 25

How do you evaluate a tensor x?

Answer 25

computed_x = x.eval(session=sess)

Question 26

How do you set the graph for a session in TensorFlow?

Answer 26

sess = tf.Session(graph=g)

Question 27

How do you close a TensorFlow session?

Answer 27

sess.close()

Question 28

How do you create a new graph in TensorFlow?

Answer 28

g_new = tf.Graph()

Question 29

How can you get the shape of a Tensor and print it to the command line as a list?

Answer 29

print(x.get_shape().as_list())

Question 30

How do you configure python to use Pyplot?

Answer 30

import matplotlib.pyplot as plt

Question 31

How do you plot inline with iPython?

Answer 31

%matplotlib inline
import matplotlib
import matplotlib.pyplot as plt

Question 32

How can you check to see if you can multiply 2 matrices?

Answer 32

There inner dimensions must match

A has m rows and n columns
B has n rows and p columns

The number of columns in A matches the number of rows in B.

Question 33

If you multiply two matrices what is the size and the resulting matrix?

Answer 33

If matrix A is m x n, and matrix B is n x p, the resulting matrix from A x B will have the dimensions m x p.

Question 34

What is convolution?

Answer 34

Convolution is a mathematical operation on two functions, to produce a third function.

Question 35

How can you create a 2-D Gaussian kernel?

Answer 35

A is a vector containing a normal distribution. Multiplying A by its transpose will give you a Gaussian kernel.

G = A * A’

Question 36

How do you multiply a matrix by its inverse in Tensorflow?

Answer 36

z = a normal distribution vector

ksize = z.get_shape().as_list()[0]

G = tf.matmul(tf.reshape(z, [ksize, 1])
tf.reshape(z, [1, ksize]))

Question 37

How do you create a Gabor kernel?

Answer 37

Modulate a Gaussian with a sine wave.

Question 38

How do you plot an image in python?

Answer 38

Use the pyplot library (matplotlib.pyplot)

plt.imshow(img)

Question 39

How do you load an image in python?

Answer 39

img = plt.imread(‘path_to_file’)

Question 40

How can you convert a list into an array in python?

Answer 40

data = np.array(list)

Question 41

How do you get the mean and standard deviation of a set of images?

Answer 41

Combine them into an array of image x height x width x channel

mean_img = np.mean(data, axis=0)
std_img = np.std(data, axis=0)

Question 42

How can you flatten an array into a vector in Python?

Answer 42

flattened = data.ravel()

Question 43

What are the six key ideas behind Stoicism?

Answer 43

1. Appreciate the shortness of life
2. Live in the present
3. Don’t seek other peoples praise or approval
4. There are no good or bad events, only your perception of them
5. Identify whats in your control and whats not
6. Find happiness in virtue

Question 44

What are the four stoic virtues?

Answer 44

• Wisdom
• Courage
• Justice
• Temperance

Question 45

What are the questions for Seneca’s evening review?

Answer 45

What bad habit did I curb today?
How am I better?
Were my actions just?
How can I improve?

Question 46

What is the difference between the L1-norm and the L2-norm?

Answer 46

These are the ‘Least Absolute Deviations’ and ‘Least Squares’ methods. Generally the L2-norm is the best for producing a stable solution. However, if the data has outliers that it is best to ignore a more robust solution may be found with the Least Absolute Deviations method.

Question 47

How do you do integer devision in Python?

Answer 47

Use the double slash
a = 4
b = 2
c = a // b

Question 48

What will np.arange(100) do in Python?

Answer 48

Create a vector of numbers from 0 to 99.

For a non-integer step size its better to use linespace.

Question 49

What are the 3 Gradient Decent methods?

Answer 49

1. Stochastic Gradient Descent : Uses 1 example of the training data each iteration.
2. Batch Gradient Descent : Uses all the training data examples each iteration.
3. Mini Batch Gradient Descent : uses a subset of the training data examples each iteration.