Learning rate should be model.

a$)$ increased If the cost oscillates while training a neural network

c$)$ kept constant

b$)$ decreased

d$)$ none of the above

To train a neural Network model, for categorical attributes we use $q,$

a$)$ One hot encoding

b$)$ Single input with different values

c$)$ Both a and b correct

d$)$ none of the above

To train a neural network model all the weights should be initialized to $q,$

a$)$ one

b$)$ zero

c$)$ random values

d$)$ both a and b correct

Which one of the following is not a model to learn word associations?

a$)$ Word $2$Vec

b$)$ GloVe

c$)$ FastText

d$)$ WordLogic

In image recognition problems we use CNN instead of normal deep neural networks because:

a$)CNN$ ensures regularization

b$)CNN$ converges faster

c$)CNN$ is easy to implement

d$)$ CNN utilizes spatial relationship among the pixels

Which one of the following is a sign of overfitting in a machine learning model?

a$)$ The accuracy on validation data decreases rapidly

b$)$ The accuracy on training data decreases

c$)$ The accuracy on validation data no longer increases

d$)$ The accuracy on training data improves very fast

Which one of the following activation functions is exclusively used in the final layer of a neural network?

a$)$ Sigmoid function

b$)$ ReLU function

b$)$ Tanh function

d$)$ Softmax function

Which one of the following cost functions is good for logistic regression?

a$)$ quadratic cost function

b$)$ cross entropy cost function

c$)$ mean absolute error cost function

d$)$ none of the above

To tune the hyperparameters which one of the following datasets is used?

a$)$ training data

b$)$ test data

c$)$ validation data

d$)$ all the above