ITPEC FE Subject A April 2026 Question 55

A driver assistance system that learns from numerous images to tell pedestrians from vehicles — image recognition trained on large data is the typical deep learning use.

Deep learning uses multi-layer neural networks that learn the distinguishing features from many examples instead of hand-written rules. Its accuracy grows with the amount of training data, which is exactly the "acquiring and processing numerous images" in option (b).

Answer (b)

Why not others:

• •(a) — an accelerometer firing the air bag at a collision is a fixed sensor threshold; nothing is learned
• •(c) — automatic idling stop is rule-based control under fixed conditions
• •(d) — updating software and maps over a mobile line is software distribution, not learning or inference

Key rule: Deep learning is recognisable by learning from large volumes of data, typically images, sound or text, where rules cannot practically be written by hand; fixed thresholds and scheduled updates are not learning.

Business Industry · AI-written, may contain errors