CNN Visualizing CNN Feature Maps 4

Question 1

Grad-CAM produces coarse localization maps. What is the primary reason for their low spatial resolution?

Accepted Answer

It relies on the final convolutional layer which has small spatial dimensions due to pooling/striding

Answer

Grad-CAM uses the last convolutional layer's activations, which have been spatially compressed by pooling and striding, producing coarse heatmaps that are then upsampled.

Question 2

Guided Backpropagation improves standard backpropagation visualization by doing what?

Accepted Answer

Zeroing out negative gradients at ReLU gates during the backward pass

Answer

Guided backpropagation zeros gradients that are negative either in the upstream gradient OR in the forward activation, producing sharper, cleaner saliency maps.

Question 3

What is Guided Grad-CAM?

Accepted Answer

The element-wise product of Guided Backpropagation and Grad-CAM outputs

Answer

Guided Grad-CAM combines the class-discriminative localization of Grad-CAM with the fine-grained detail of Guided Backpropagation via element-wise multiplication.

Question 4

In feature map visualization, what does 'channel-wise mean activation' summarize?

Accepted Answer

The average activation value across all spatial positions in a feature map

Answer

Channel-wise mean activation collapses the spatial dimensions by averaging all positions in a feature map, giving a single scalar representing each channel's overall response.

Question 5

Which statement about LIME (Local Interpretable Model-agnostic Explanations) in the context of CNNs is correct?

Accepted Answer

LIME perturbs superpixels of the input and fits a local linear model to explain predictions

Answer

LIME explains CNN predictions by masking image superpixels, observing prediction changes, and fitting a simple interpretable model to approximate local decision boundaries.

CNN - Convolutional Neural Networks Practice Test

CNN - Convolutional Neural Networks Practice Test

CNN Visualizing CNN Feature Maps 4