Questions: Curse of Dimensionality

5 questions to test your understanding

Score: 0 / 5
Question 1 Multiple Choice

A team trains a k-nearest neighbors classifier on 500 medical records with 5 features, achieving 82% test accuracy. They add 300 additional measured features to the same 500 records and retrain, achieving 99% training accuracy but 68% test accuracy. What is the most likely explanation?

AThe 300 new features are noisy and should have been normalized before training
Bk-NN requires recalibrating k when the number of features changes, and the team used the wrong k
CIn high-dimensional space the data is too sparse for meaningful distance comparisons, and the model exploits spurious patterns — a manifestation of the curse of dimensionality
D500 records is a large enough dataset for any number of features; the problem is overfitting due to wrong hyperparameters
Question 2 Multiple Choice

As the number of dimensions in a dataset increases, what happens to the ratio of nearest-neighbor distance to farthest-neighbor distance from a reference point?

AThe ratio increases — nearest neighbors get much closer relative to farthest neighbors
BThe ratio stays constant — distances scale uniformly in all directions
CThe ratio approaches 1 — all points become approximately equidistant from the reference
DThe ratio approaches 0 — the farthest neighbors become infinitely distant
Question 3 True / False

Adding more features to a machine learning model with a fixed training set usually improves or maintains test accuracy, since additional features provide the model with more information.

TTrue
FFalse
Question 4 True / False

For distance-based algorithms like k-nearest neighbors, high dimensionality can make the 'nearest neighbor' concept meaningless because all pairwise distances between points become approximately equal.

TTrue
FFalse
Question 5 Short Answer

Why does adding features to a fixed-size dataset make it harder for a model to generalize, even if those features carry real signal?

Think about your answer, then reveal below.