Question: In a concise analytical response, explain how differences between internet and mobile-app images can create a representation-capacity tradeoff and affect the target task.

Sample answer: Internet images may have source-specific properties, such as higher resolution or different framing. A neural network must expend some representational capacity to learn these properties. If they differ greatly from mobile-app images, less capacity remains for recognizing the mobile-app distribution that matters, so target performance could theoretically decline.

Key points:

• Internet images may differ in resolution or framing.
• The network spends capacity learning source-specific properties.
• Large source differences can leave less capacity for mobile-app images.
• Reduced target-focused capacity could hurt performance.

Rubric: A strong response identifies a source-specific property, explains that learning it consumes representational capacity, connects this consumption to reduced capacity for mobile-app recognition, and states the possible performance consequence.