quiz bu lëkkale ak tegtal · Diggante Niveau

The Curse of Dimensionality Quiz

Tests sparsity, distance concentration, nearest-neighbor limits, sample-size effects, practical strategies for high-dimensional data, and validation.

Yooni njiit yu jëm ciCurse Of Dimensionality
Laaj 1 de 8

What does the curse of dimensionality describe?

Xoolaatal laaj bu nekk8 laaj

Sooy def quiz bi, tontu yi ak leeral yi dañuy feeñ.

  1. What does the curse of dimensionality describe?
    • A rule that all high-dimensional data must be reduced to two variables
    • A type of label error caused by imbalanced classes
    • Challenges in data coverage, distance behavior, or computation as the number of dimensions grows
    • A guarantee that every model fails above a fixed feature count
  2. Why can a fixed-size dataset become sparse in high dimensions?
    • More dimensions create a larger feature space to cover with the same observations
    • The sample size automatically decreases when columns are added
    • High dimensions remove all correlations between features
    • Each feature is forced to contain fewer unique values
  3. What can happen to nearest and farthest distances as dimension grows?
    • Distance becomes a direct measure of causal influence
    • Their contrast can shrink, making distance less discriminating
    • The nearest distance always becomes exactly zero
    • The farthest distance is always smaller than every nearest distance
  4. Does the curse of dimensionality mean every high-dimensional model fails?
    • Yes; every model fails after ten dimensions
    • Yes; adding any feature reduces accuracy
    • No; high-dimensional models always outperform simple ones
    • No; effects depend on data structure, method, sample size, and noise
  5. Which methods can help with irrelevant or redundant features?
    • Feature selection, regularization, or a justified lower-dimensional representation
    • Adding more unrelated features to increase coverage
    • Randomly scaling each feature until distances change
    • Tuning on the test set until validation improves
  6. Why fit feature selection using only training data?
    • To make test observations independent of the outcome
    • To maximize the apparent training accuracy
    • To prevent information from the holdout set leaking into model development
    • To ensure every feature is kept in the final model
  7. What does approximate nearest-neighbor search trade?
    • It makes the dataset lower-dimensional without transformation
    • It can improve search efficiency while allowing results to differ from exact neighbors
    • It guarantees that classification accuracy improves
    • It removes the need for a distance metric
  8. What should a report about a high-dimensional model specify?
    • Only the number of input columns
    • The single best training score
    • The name of the dimensionality-reduction package
    • Sample size, preprocessing, distance or representation choices, and validation design