Final (3) Flashcards
(17 cards)
What does NB assume about features?
They are independent given the class
What does BernoulliNB do?
Checks if features are 0 or non-zero for each class.
What does MultinomialNB use?
Average values of features per class.
What does a large alpha in NB do?
Adds smoothing and reduces model complexity.
NB model strengths?
Fast to train and predict, similar pros/cons to linear models.
What do hidden layers allow neural nets to do?
Learn complex, non-linear decision boundaries.
What do activation functions do?
Add non-linearity to hidden layers.
Neural net strengths?
Can capture patterns in large datasets and beat other models with tuning.
Neural net weaknesses?
Long training time, needs preprocessing, works best with homogenous data.
What is GridSearchCV used for?
Finding the best model parameters through exhaustive search.
What does a Pipeline do?
Combines preprocessing and modeling steps into one object.
Why do we use cross-validation?
To get a reliable average accuracy score and avoid information leakage.
What does the R² score measure?
Goodness of fit.
What makes a good model in terms of bias and variance?
Low bias (train accuracy close to 1), low variance (train ≈ test).
What is the goal of k-means clustering?
Minimize within-cluster sum of squares (inertia).
How does Agglomerative Clustering work?
Starts with each point as its own cluster and merges the most similar ones.
What does DBSCAN look for?
Dense regions in feature space.
