AI and machine learning interview questions

MAE averages absolute errors and treats misses in a linear way.

Linear regression predicts a continuous target as a weighted sum of features plus an intercept.

It assumes a roughly linear relationship, independent errors, and fairly constant residual variance.

A residual is actual y minus predicted y for a row.

Holding other features fixed, the coefficient is the expected change in y for a one-unit rise in that feature.

Logistic regression models class probability with a linear score passed through a sigmoid.

A linear fit can predict values far outside 0 and 1 and is sensitive to extreme labels.

The sigmoid maps any real number onto a probability between 0 and 1.

A coefficient is the change in log-odds for a one-unit feature increase with other features held fixed.

The tree splits the feature space with if-then rules that improve purity or reduce error at each node.

Gini impurity measures how mixed the classes are inside a node.

Information gain is the drop in entropy after a split.

A random forest is many decision trees trained on bootstrap samples of the rows.

One deep tree overfits

Bagging trains the same algorithm on bootstrap resamples and then aggregates the predictions.

Boosting builds models in sequence so each new model focuses on leftover errors.

Bagging trains models in parallel on random subsets and averages them.

XGBoost and LightGBM are the names interviewers hear most, with CatBoost close behind.

An SVM finds a decision boundary that maximizes the margin between classes.

Support vectors are the training points that sit on or inside the margin.