IPA IP 2022 Question 67

A neural network that learns patterns from large volumes of data — deep learning finds the patterns by itself and uses them to infer or decide.

Deep learning is a form of machine learning that uses a neural network with many layers. Trained on a large volume of data, the network adjusts its internal weights so that successive layers extract increasingly abstract features, without people writing the rules or choosing the features. It underlies image recognition, speech recognition and machine translation.

Answer (b)

Why not others:

• •(a) — describes e-learning: studying at one's own level from materials on the Internet, anywhere and at any time
• •(c) — describes an expert system, which answers from specialist knowledge that people have organised into rules
• •(d) — describes adaptive learning, which sets the difficulty for each learner from their proficiency and understanding

Key rule: Deep learning is machine learning with multilayer neural networks that learns features from large data; an expert system instead follows rules written down from specialists' knowledge.

Basic Theory · AI-written, may contain errors