Data Science and Machine Learning (Theory and Projects) A to Z - Deep Neural Network Architecture: Filters Padding Strid

The video tutorial explains the concept of filters and masks in convolutional neural networks (CNNs), detailing how filter banks are used to process input structures. It covers the application of filters, the role of activation functions and bias, and the importance of padding and stride in determining output dimensions. The tutorial emphasizes the flexibility of these parameters and their impact on the CNN's performance.

What is another term for a convolution mask?

• Filter bank

• Kernel

• Matrix

• Tensor

What is the typical relationship between the number of channels in a filter and the input structure?

• The input structure has more channels

• They have the same number of channels

• The filter has fewer channels

• The filter has more channels

In a CNN, what is the purpose of adding a bias before applying an activation function?

• To reduce computation time

• To increase the size of the output

• To adjust the filter size

• To shift the activation function

What is the main reason for using padding in convolutional layers?

• To change the filter size

• To increase the number of filters

• To enhance image quality

• To prevent boundary issues

Which padding method involves mirroring the image across a boundary?

• Constant padding

• Zero-padding

• Symmetric reflectance

• Reflective padding

Which of the following is NOT a method of padding?

• Random padding

• Constant padding

• Symmetric reflectance

• Zero-padding

What does the stride parameter control in a convolutional layer?

• The step size of the filter movement

• The number of channels

• The size of the input image

• The number of filters

How does increasing the stride value affect the output dimensions?

• Increases the output size

• Doubles the output size

• Decreases the output size

• Has no effect on the output size

What is the effect of setting a stride value greater than one?

• Has no effect on the output

• Increases the input dimensions

• Reduces the output dimensions

• Increases the number of filters

What is a hyperparameter in the context of CNNs?

• A fixed value for all layers

• A parameter that can be learned from data

• A parameter set before training

• A parameter that changes during training