Deep Learning Convolutional Neural Networks 2

Question 1

What is the effect of adding zero-padding to the input of a convolutional layer?

Accepted Answer

Preserves spatial dimensions by adding zeros around the input border

Answer

Zero-padding adds zeros around the input border so that the output feature map maintains the same spatial size as the input after convolution.

Question 2

In transfer learning with CNNs, what are the early layers of a pretrained model typically found to detect?

Accepted Answer

Low-level features such as edges and textures

Answer

Early CNN layers learn generic low-level features like edges, corners, and color gradients that are transferable across image recognition tasks.

Question 3

What is a depthwise separable convolution, as used in MobileNet?

Accepted Answer

A factored convolution splitting spatial and channel mixing into separate steps to reduce computation

Answer

Depthwise separable convolutions apply a single filter per input channel (depthwise) then combine channels with a 1×1 convolution (pointwise), greatly reducing FLOPs.

Question 4

What problem does the Inception module in GoogLeNet address?

Accepted Answer

Choosing filter size by applying multiple filter sizes in parallel and concatenating outputs

Answer

The Inception module applies 1×1, 3×3, and 5×5 convolutions in parallel and concatenates results, allowing the network to capture multi-scale features automatically.

Question 5

What is the main purpose of 1×1 convolutions in deep CNN architectures?

Accepted Answer

Perform channel-wise linear combinations to reduce or expand depth cheaply

Answer

1×1 convolutions mix information across channels without affecting spatial dimensions, enabling dimensionality reduction (bottleneck) or expansion at low computational cost.

Deep Learning Practice Test

Deep Learning Practice Test

Deep Learning Convolutional Neural Networks 2