week 4

What is the main difference between Logistic Regression and Softmax Regression?

• A) Logistic regression is used for multi-class classification, while Softmax is for binary classification.

• B) Logistic regression is used for binary classification, while Softmax is for multi-class classification. (Correct Answer)

• C) Logistic regression requires a Softmax function to work.

• D) There is no difference; they are the same.

Which of the following correctly describes the shape and dimensions of grayscale and colored images?

• Grayscale images have one channel (H x W x 1), while colored images have three channels (H x W x 3).

• Colored images have a single-channel representation.

• Both grayscale and colored images have three channels.

• Grayscale images cannot be processed by CNNs.

What is the purpose of a Flatten Layer?

• It combines the output from different CNN models.

• It converts the 2D feature maps into a 1D vector before feeding into fully connected layers. (Correct Answer)

• It removes unnecessary details from feature maps.

• It increases the number of layers in the network.

Can deep MLPs solve the CIFAR-10 dataset problem efficiently? Why or why not?

• No, deep MLPs struggle with spatial features in images. CNNs perform better as they preserve spatial hierarchies and learn local patterns using convolutional layers

• Yes, they are specifically designed for image processing tasks.

• Yes, they use convolutional layers to process images.

• Yes, they are computationally inexpensive.

How much do you already know about CNNs? (Select one)

• First time hearing about them

• A little knowledge

• Average knowledge

• Good knowledge

What is the role of a Kernel/Filter in CNNs?

• To extract spatial features from images

• To increase the image resolution

• To directly classify the image

• To store all pixel values for learning

What is the purpose of Stride and Padding in CNNs?

• Stride controls how the filter moves across the input image, affecting the output size. Padding helps preserve the spatial size of the input by adding borders around the image.

• Stride helps in preserving the spatial size of the input, while Padding controls the filter movement.

• Both Stride and Padding are used to increase the depth of the CNN.

• Stride and Padding have no specific purpose in CNNs.

 Which operation is commonly used to reduce the spatial dimensions of feature maps in CNNs?

• Flattening

• Pooling

• Convolution

• Activation

Why is shuffling data important before training a machine learning model?

• It helps to improve the performance of the model

• It ensures that the model sees the data in a random order

• It prevents overfitting of the model

• It reduces the computational complexity of the training process

 Which common challenge in training deep neural networks is characterized by gradients becoming extremely small, leading to slow or stalled learning?

• Overfitting

• Vanishing gradients

• Exploding gradients

• Underfitting