Which of these are reasons for Deep Learning recently taking off

• We have access to a lot more computational power

• Neural networks are a brand new field

• We have access to lot more data

• Deep learning models need less computational power

If you increase the number of hidden layers in a multi-layer Perceptron, the classification error of test data always decreases.

• True

• False

In a neural network, which of the following techniques is used to deal with overfitting?

• Dropout

• Regularization

• Batch normalization

• All of these

For an image recognition problem (recognizing a cat in a photo), which architecture of neural network would be better suited to solve the problem?

• Multi-layer perceptron

• Convolutional neural network

• Recurrent neural network

• All of these

Select non-linear activation functions from below

• ReLU

• Tanh

• Binary step

• Adagrad

Which of the below are optimization functions

• Adam

• Adagrad

• Stochastic gradient descent

• All of the above

Recommender system which relies on past user-item interactions to make new predictions is

• Content-based

• Collaborative filtering

Recommender system approach which learns from existing features to explain user-item is

• Content-based

• Collaborative filtering

Which of the following is not a part of text normalization process

• Stop word removal

• Text tokenization

• Document Labelling

• Lemmatization

TF-IDF penalizes heavily words that occur very frequently in a document

• True

• False