Questionnaire 5

JavaScript isn't enabled in your browser, so this file can't be opened. Enable and reload.

Questionnaire 5

Email *

Record my email address with my response

Full name
*

Group
*

001

002

003

004

arm

Question 1 *

Select the correct statements about neural networks.

Neural networks with multiple hidden layers allow modeling complex feature interactions.

Increasing the number of neurons in a hidden layer enhances the neural network's ability to model complex nonlinear dependencies.

Increasing the number of hidden layers enhances the neural network's ability to model complex nonlinear dependencies.

Neural networks are not suitable for multiclass classification tasks.

Required

Question 2 *

Select all the correct statements about activation functions.

Softmax is usually used to activate the output layer in multiclass classification.

Sigmoid can be used to activate the output layer in binary classification.

The linear activation function is used for regression tasks, in the output layer of a neural network.

The ReLU activation is intended for neurons in the internal layers.

Required

Question 3
*

Which activation function is recommended to use in the hidden layer of a neural network?

ReLU

Tanh

Linear

Sigmoid

Question 5
*

How many neurons should the output layer of a neural network designed for binary classification contain?

Question 6 *

How many neurons will the output layer of a neural network designed for multi-class classification task with 7 classes contain?

Question 7
*

Having a neural network with an input layer of 1000 neurons, one hidden layer of 500 neurons, and an output layer of 10 neurons, which of the following initialization methods is better to use for the weights connecting the input and hidden layers?

Random initialization from a uniform distribution with a range of (-0.06, 0.06).

Random initialization from a uniform distribution with a range of (0, 0.1).

Initialization of all weights with the same value from the range (-0.063, 0.063).

Random initialization from a uniform distribution with a range of (-0.141, 0.141).

Question 8
*

Choose the correct statements about training neural networks.

Dropout helps prevent overfitting.

Neurons that were excluded from computations due to dropout during training are remembered and also excluded during testing of the network.

Dropout is the only way to regularize neural networks.

Dropout is for reducing the training time.

Required

Question 9
*

In backpropagation, what is the role of the gradient descent algorithm?

To compute the gradients of the loss function with respect to the weights

To update the weights of the neural network based on the computed gradients

To randomly drop out neurons during training to improve performance

Send me a copy of my responses.

Submit

Clear form

Never submit passwords through Google Forms.

This form was created inside of Russian-Armenian (Slavonic) University. Report Abuse

Forms