Which of the following options is not a reason for traditional machine learning algorithms to promote the development of deep learning?

Question

Options

• ADimensional disaster
• Blocal invariance and smooth regularization
• CManifold learning
• DFeature Engineering

• A
  9% (3)
• B
  16% (5)
• C
  3% (1)
• D
  72% (23)

Explanation

Feature Engineering (D) is the correct answer because it describes a technique used by traditional ML, not a limitation of it. Deep learning was developed precisely to replace manual feature engineering with automatic feature extraction - so it cannot simultaneously be a reason traditional ML promoted deep learning's development.

Why the distractors are wrong (i.e., A, B, C are genuine reasons):

• A (Curse of dimensionality): Traditional ML degrades sharply as input dimensions grow, motivating deep learning's ability to learn compact representations in high-dimensional spaces.
• B (Local invariance and smooth regularization): Traditional ML heavily assumes locally smooth decision boundaries, which breaks down on complex, non-smooth real-world data like images and speech - a gap deep learning fills.
• C (Manifold learning): Traditional ML struggles to discover the low-dimensional manifold structure hidden in high-dimensional data; deep networks learn hierarchical manifold representations implicitly.

Memory tip: Think of A, B, C as theoretical ceilings that traditional ML hit - deep learning was built to break through them. D is the opposite: it's a burden traditional ML imposed that deep learning removed. Anything deep learning eliminates cannot be what motivated its creation.

No community discussion yet for this question.