Deep Learning, Quiz-2

Define filters in the context of CNN layers.

• Filters in CNN layers are applied only once to the input data

• Filters in CNN layers are small grids applied to input data to extract specific features by sliding over the input data and performing element-wise multiplication and summation to produce feature maps.

• Filters in CNN layers are used to resize images

• Filters in CNN layers are used to adjust brightness levels of images

Describe the function of a fully connected layer in CNN.

• Fully connected layers are only applied to the input layer in CNNs

• Fully connected layers are only used in RNNs, not CNNs

• The function of a fully connected layer in CNN is to connect every neuron in one layer to every neuron in another layer, enabling complex relationships to be learned.

• Fully connected layers are used to downsample images in CNNs

How does L1 regularization differ from L2 regularization?

• L1 regularization is more computationally efficient than L2 regularization.

• L1 regularization adds the squared values of the coefficients, while L2 regularization adds the absolute values of the coefficients.

• L1 regularization adds the absolute values of the coefficients to the loss function, while L2 regularization adds the squared values of the coefficients.

• L1 regularization does not penalize large coefficients, unlike L2 regularization.

Explain the concept of data augmentation in the context of CNN.

• Data augmentation in CNN is only applicable to testing data, not training data.

• Data augmentation in CNN involves applying transformations to training data to create variations, increasing diversity and preventing overfitting.

• Data augmentation in CNN involves reducing the diversity of training data to improve model performance.

• Data augmentation in CNN has no impact on preventing overfitting.

What is the impact of increasing the number of filters in a CNN layer?

• Increasing the number of filters enhances the model's ability to extract higher-level features from the input data.

• Increasing the number of filters leads to overfitting in the model.

• Increasing the number of filters has no impact on the model's performance.

• Increasing the number of filters reduces the model's ability to learn complex patterns.

What is the purpose of activation functions in CNN?

• Activation functions are used to increase the size of the input data

• Activation functions are applied to reduce the complexity of the model

• Activation functions introduce non-linearity to the network, allowing it to learn complex patterns and make predictions.

• Activation functions are used to skip the convolutional layers

Explain the concept of feature maps in CNN.

• Feature maps are used to store the original input data

• Feature maps are visual representations of the input data

• Feature maps in CNN are the output of applying filters to the input data, highlighting specific features or patterns.

• Feature maps are only used in the output layer of CNN

What is the role of stride in convolutional layers of CNN?

• Stride is used to increase the size of the filters

• Stride determines the number of filters applied to the input data

• Stride controls the amount of movement of the filter over the input data, affecting the size of the output feature map.

• Stride is used to decrease the number of layers in the CNN

Explain the concept of learning rate in the context of training deep neural networks.

• Learning rate is the number of layers in a neural network

• Learning rate is the rate at which the model learns from the training data by updating the weights during backpropagation.

• Learning rate is the size of the input data

• Learning rate is the number of neurons in a neural network

Describe the role of epochs in training a deep neural network.

• Epochs determine the number of layers in the neural network

• Epochs control the learning rate of the model

• Epochs are used to adjust the dropout rate in the model

• Epochs represent the number of times the entire training dataset is passed forward and backward through the neural network.