DL_Unit-3

What aspect of CNNs is focused on preventing overfitting during training?

• • Dropout layers

• • Batch normalization

• • Sigmoid activation

• • ReLU activation

In the context of CNNs, what does the term "underfitting" indicate?

• High training loss

• Insufficient model complexity

• Perfect model performance

• Overemphasis on training data

In CNNs, what does the term "Multiple Filters" refer to?

• Various image sizes

• Different convolutional layers

• Various feature detectors

• Diverse activation functions

When encountering the problem of underfitting in CNNs, what advanced technique can be employed to enhance model performance?

• Ensemble learning

• Transfer learning

• • Hyperparameter tuning

• • Quantum computing

In the context of CNNs, what does the term "underfitting" indicate?

• Insufficient model complexity

• Perfect model performance

• Overemphasis on training data

• High training loss

What is the purpose of the Batch Normalization layer in CNNs?

• Enhancing model complexity

• Accelerating convergence

• Normalizing input data for each batch

• Reducing the number of layers

Complete Roll Number

In CNNs, what is the role of the Global Average Pooling (GAP) layer in the feature extraction process?

• Reducing model complexity

• Summarizing spatial information

• Increasing filter size

• Ignoring pooling operation

What happens to the receptive field of a neuron in a CNN when using a larger convolutional filter?

• It remains unchanged

• It decreases

• It increases

• It depends on the stride value

What role does zero-padding play in convolutional filters in a CNN?

• Enhances filter visualization

• Prevents loss of spatial information

• Increases computational cost

• Reduces filter diversity