GATE DA Unsupervised Learning & Dimensionality Reduction – Data Science & AI

GATE DA Unsupervised Learning & Dimensionality Reduction – Data Science & AI | Set 1

Questions 41-45 of 100

0% completed (0/100 answered)

Question 41

MCQ · Single Correct 2 mark

Kernel PCA extends standard linear PCA to perform non-linear dimensionality reduction by:

A Clustering data into manifold slices B Scaling features logarithmically C Using non-linear decision trees before projecting data D Applying the Kernel Trick to compute inner products in a high-dimensional feature space $\Phi(x)$

Question 42

MCQ · Single Correct 2 mark

t-SNE (t-Distributed Stochastic Neighbor Embedding) is primarily used for:

A Supervised feature selection for linear models B Exact reconstruction of missing data C Low-dimensional visualization of high-dimensional non-linear manifolds D Linear classification of separable data

Question 43

MCQ · Single Correct 2 mark

Why does t-SNE use a Student-t distribution (with 1 degree of freedom / Cauchy distribution) in the low-dimensional map instead of a Gaussian distribution?

A To alleviate the 'crowding problem' by allowing moderately distant points to be mapped farther apart B To enforce strict linear projection boundaries C Because Student-t distribution is faster to compute than Gaussian D Because Gaussian distribution is non-differentiable

Question 44

MCQ · Single Correct 2 mark

What loss function does t-SNE minimize during optimization?

A Binary Cross-Entropy B Hinge Loss C Mean Squared Error (MSE) D Kullback-Leibler (KL) Divergence KL(P || Q)

Question 45

MCQ · Single Correct 2 mark

In Multidimensional Scaling (MDS), what property of the data is preserved in the lower-dimensional representation?

A Pairwise distances / dissimilarity matrix between data points B Class label separation boundaries C Decision tree split thresholds D Cluster variance ratios