In SAS DLPy, a data scientist trains a CNN model and wants to visualize which regions of an input image most influenced the model's prediction. Which technique is commonly applied for this purpose?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — you want Class Activation Mapping (CAM) or Gradient-weighted Class Activation Mapping (Grad-CAM) because it creates heatmaps showing which image regions most influenced a class prediction. It works by relating the class score to the final convolutional feature maps, and SAS DLPy supports activation map visualization.
Full Explanation
Class Activation Mapping (CAM) or Gradient-weighted Class Activation Mapping (Grad-CAM) is the appropriate technique because it generates heatmaps showing which spatial regions of the input image contributed most to a specific class prediction by computing the gradient of the class score with respect to the final convolutional feature maps. SAS DLPy supports activation map visualization. The PCA projection of the convolutional feature maps onto the input image space is a dimensionality reduction technique, not a saliency method. The t-SNE visualization of the final fully connected layer activations visualizes latent space distributions, not input image relevance. The PROC FACTOR rotation applied to the convolutional kernel weights operates on statistical variables, not neural network visualizations.