Tensorflow on Google Cloud

Question 1

What is a Tensor?

Answer 1

Tensor represents a n-dimensional array of data (scalar, vector, matrix, n-dimensional matrix)

Question 2

Explain Tensorflow layering architecture

Answer 2

It consists of multiple layers, from the lowest level in C++, over Python, Python modules and in the end high level python functions.

Question 3

How Tensorflow represent numerical operations?

Answer 3

Using a Directed Acyclic Graph (DAG)

Question 4

What is the difference between tf.constant and tf.Variable

Answer 4

tf.constant - creates a constant tensor that can’t be changed
tf.Variable - creates a tensor that can change its value

Question 5

What is Keras?

Answer 5

Keras is a high-level API that enables you to work with Tensorflow to load data, train and evaluate models, create layers, etc. It represents an abstraction layer that is much easier to use.

Question 6

What are some basic methods working with models?

Answer 6

model.compile() - define learning rate, loss function, optimizer…
model.fit() - start the training and provide additional things like number of epohs, train and test dataset
model.evaluate() - get model metrics

Question 7

What is tf.data API?

Answer 7

It is used to create data pipelines in Tensorflow. You can read data, do some transformations, create features and many more.