Testy Fiszki Notatki

Zaloguj

Fiszki

DeepLearning_GPT3_questions

Test w formie fiszek

Ilość pytań: 85 Rozwiązywany: 610 razy

What is the purpose of the softmax function in deep learning?

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

To normalize the output of the neural network to a probability distribution

To calculate the output of the neural network

To activate the neurons in the neural network

To normalize the output of the neural network to a probability distribution

What is the purpose of the backpropagation algorithm in deep learning?

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

To update the weights in the neural network

To calculate the output of the neural network

To propagate the input forward through the network

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

What is the difference between supervised and unsupervised learning?

Supervised learning requires labeled data, while unsupervised learning does not

Supervised learning is more accurate than unsupervised learning

There is no difference between the two

Supervised learning requires less training data than unsupervised learning

Supervised learning requires labeled data, while unsupervised learning does not

Which of the following is not a commonly used activation function in deep learning?

Sigmoid

Polynomial

ReLU

Tanh

Polynomial

What is the purpose of regularization in deep learning?

To reduce the variance in the training data

To increase the accuracy of the model

To reduce the bias in the model

To prevent overfitting

What is the purpose of using dropout in convolutional neural networks?

To increase the number of parameters in the network

To increase the accuracy of the model

To reduce the computational cost of the model

To prevent overfitting in the model

Answer: b) To prevent overfitting in the model

Explanation: Dropout is a regularization technique used in convolutional neural networks to prevent overfitting in the model. Dropout randomly drops out a certain percentage of the neurons in the network during training, forcing the network to learn more robust features. This helps prevent the model from memorizing the training data and improves its generalization performance on unseen data.

What is the purpose of using padding in convolutional neural networks?

To increase the number of filters in the convolutional layer

To reduce overfitting in the model

To reduce the spatial dimensions of the input volume

To ensure that the output volume has the same spatial dimensions as the input volume

c) To ensure that the output volume has the same spatial dimensions as the input volume

Explanation: Padding is used in convolutional neural networks to ensure that the output volume has the same spatial dimensions as the input volume. This is necessary when using convolutional layers without any pooling layers, since the output size would be reduced after each convolutional layer.

Which of the following is used to reduce the spatial dimensions of the input volume in a convolutional neural network?

Convolutional layers

Fully connected layers

Pooling layers

Activation functions

Pooling layers

) Pooling layers

Explanation: Pooling layers are used to reduce the spatial dimensions of the input volume in a convolutional neural network. Max pooling and average pooling are the two commonly used types of pooling layers.

What is the output shape of a convolutional layer with 32 filters, a filter size of 3x3, and input shape of 224x224x3?

222x222x32x3

32x32x3

222x222x32

222x222x3

222x222x32

b) 222x222x32

Explanation: The output shape of a convolutional layer can be calculated using the formula: (W - F + 2P)/S + 1, where W is the input shape, F is the filter size, P is the padding size, and S is the stride. Applying this formula, we get (224 - 3 + 2*0)/1 + 1 = 222. Therefore, the output shape is 222x222x32.

Which of the following statements is true about convolutional neural networks?

Convolutional neural networks cannot be used for object detection tasks.

Convolutional neural networks can only be used for image classification tasks.

Convolutional layers are always followed by fully connected layers.

The use of convolutional layers reduces the number of parameters in the network.

b) The use of convolutional layers reduces the number of parameters in the network.

Explanation: Convolutional neural networks are used for various tasks including image classification, object detection, and segmentation. The use of convolutional layers reduces the number of parameters in the network by sharing weights across different locations in the input, making the model more efficient.

Which of the following is a common technique used to prevent overfitting in convolutional neural networks?

All of the above

Dropout

Early stopping

Data augmentation

All of the above

Explanation: Overfitting is a common problem in convolutional neural networks, and several techniques can be used to prevent it, including early stopping, dropout, and data augmentation. Early stopping involves monitoring the performance of the network on a validation set and stopping training when the validation error stops improving. Dropout involves randomly dropping out some of the neurons in the network during training, which can help prevent over-reliance on particular features. Data augmentation involves artificially increasing the size of the training set by applying random transformations to the input data.

What is the main advantage of using convolutional layers in CNNs?

To introduce nonlinearity into the network

To increase the number of learnable parameters

To increase the receptive field size of the network

To reduce the spatial resolution of the input

To increase the receptive field size of the network

Answer: b) To increase the receptive field size of the network

Explanation: Convolutional layers are used in CNNs to increase the receptive field size of the network, allowing the network to capture larger spatial structures in the input data. This can help improve the network's ability to classify objects in the input data.

Which of the following is a common activation function used in convolutional neural networks?

Tanh

Softmax

ReLU

Sigmoid

ReLU

b) ReLU

Explanation: Rectified Linear Units (ReLU) is a common activation function used in convolutional neural networks. ReLU is computationally efficient, easy to optimize, and has been shown to work well in practice.

What is the main advantage of using pooling layers in CNNs?

To increase the receptive field size of the network

To increase the spatial resolution of the feature maps

To reduce the number of learnable parameters

To introduce nonlinearity into the network

To reduce the number of learnable parameters

Answer: a) To reduce the number of learnable parameters

Explanation: Pooling layers are used in CNNs to reduce the spatial resolution of the feature maps, thereby reducing the number of learnable parameters in the network. This can help prevent overfitting, improve computational efficiency, and allow the network to generalize better to unseen data.

Which of the following is a common problem in image classification that convolutional neural networks (CNNs) aim to address?

Underfitting on large datasets

Slow training times

Lack of interpretability of the models

Overfitting on small datasets

Answer: a) Overfitting on small datasets

Explanation: Overfitting is a common problem in image classification when training deep neural networks on small datasets. CNNs are designed to address this issue by introducing parameters that are shared across the image, reducing the number of learnable parameters and allowing the network to generalize better to unseen data.

What is dropout regularization?

It adds the sum of absolute values of the weights as a penalty term to the loss function.

It adds a Gaussian noise term to the weights during training.

It adds the sum of squared values of the weights as a penalty term to the loss function.

It randomly removes a fraction of the neurons from the network during training.

What is L2 regularization?

It adds the sum of squared values of the weights as a penalty term to the loss function.

It adds the maximum squared value of the weights as a penalty term to the loss function.

It adds the sum of absolute values of the weights as a penalty term to the loss function.

It adds the maximum absolute value of the weights as a penalty term to the loss function.

It adds the sum of squared values of the weights as a penalty term to the loss function.

What is the effect of increasing the regularization parameter in L2 regularization?

It reduces the number of non-zero weights.

It reduces the magnitude of the weights.

It increases the magnitude of the weights.

It has no effect on the weights.

It reduces the magnitude of the weights.

What is weight decay?

It adds a Gaussian noise term to the weights during training.

It stops training the network when the validation error stops decreasing.

It adds the sum of absolute values of the weights as a penalty term to the loss function.

It adds the sum of squared values of the w loss function.

D. Weight decay adds the sum of squared values of the weights multiplied by a regularization parameter as a penalty term to the loss function.

Which of the following is a technique used for data augmentation as a regularization technique?

Subtracting noise from the input data

Removing a random subset of the input data

None of the above

Adding noise to the input data

Początek Pokaż poprzednie pytania Pokaż kolejne pytania

Powiązane tematy

Inne tryby

Nauka Test Powtórzenie