All MCQs

Multiplying by a zero likelihood makes the entire class posterior zero.
Smoothing assigns small nonzero probabilities to unseen events.

Bernoulli Naive Bayes models each feature as a binary event.
It is useful when presence and absence both carry information.

Multinomial Naive Bayes models count or frequency data.
It is widely used for document and text classification.

Gaussian Naive Bayes estimates a mean and variance per feature and class.
It evaluates each continuous value using a normal density.

Posterior probabilities are compared across classes.
The maximum a posteriori class is selected.

Likelihood measures how compatible an observation's features are with a class.
It is combined with the prior to form a posterior.

Class priors often come from class frequencies in training data.
They represent baseline class prevalence.

Bayes' theorem combines prior class probability with feature likelihood.
The result is a posterior probability for each class.

Naive Bayes factorizes the class-conditional likelihood across features.
The independence assumption is often unrealistic but computationally useful.

A modest change can alter the best root split and cascade through the tree.
Ensemble methods reduce this instability.

Limiting depth restricts model complexity.
A shallow tree may generalize better but can underfit.

Larger leaf-size requirements reduce highly specific partitions.
This acts as a regularization parameter.