AI and machine learning interview questions

A smaller feature subset at each split makes trees more diverse and usually lowers correlation among them.

Impurity importance favors high-cardinality and correlated features.

Extremely randomized trees pick split thresholds more randomly, not just the feature subset.

Gradient boosting fits each new tree to the gradient of the loss with respect to the current ensemble prediction.

The learning rate shrinks each new tree's contribution so later trees correct more gently.

AdaBoost reweights training rows so misclassified points get more attention in the next weak learner.

Boosting keeps chasing remaining train error, so extra rounds can fit noise.

C trades off margin width against training mistakes.

Both are linear classifiers.

RBF SVMs scale badly with n because they depend on many support vectors.

Classification votes among the K neighbors.

In high dimensions, distances concentrate and nearest neighbors stop being truly near.

Closer neighbors can vote more than far ones, which softens the effect of a clumsy K.

Silhouette compares how close a point is to its own cluster versus the next nearest cluster.

DBSCAN grows clusters from dense neighborhoods and labels sparse points as noise.

Inertia always falls as K grows, even when extra clusters are meaningless.

Use internal scores such as silhouette or Davies-Bouldin, plus visual checks on a 2D projection.

Look at the explained-variance ratio and keep components until a target such as 90 or 95 percent is reached.

PCA is better when many correlated numeric features share a low-dimensional subspace and you mainly need compression.

Trees split on original features and can ignore irrelevant axes without rotation.